Expand LSP capabilities, add full Go ASN.1 frontend, and broaden CI coverage

[rafael2knokia]

Acknowledgement

This PR was inspired in significant part by
Vanadium by Mikhail Krylov
(BSD-3-licensed). The pure-Go ASN.1 frontend, the Wadler-style
formatter combinator layer, the schema-driven TTCN-3 AST generator
and the lint Rule lifecycle all follow Vanadium's design. The
attribution that should have shipped with the original commit is
now in THIRD_PARTY_NOTICES.md, in per-package
doc comments, and in the README's new Acknowledgements section.

No verbatim Vanadium source code is included; the ports are
reimplementations in Go.

Summary

This PR is a broad pass over the LSP, formatter, analysis, project,
and CI layers. It is split into self-contained commits so they can
be reviewed (or reverted) independently.

LSP

• Wire the existing CLI lint engine and a new semantic analyzer into
  textDocument/publishDiagnostics.
• Add a textDocument/codeAction handler that consumes autofix payloads
  produced by the linter, and emit a source.organizeImports action
  that sorts, dedupes, and respects visibility per module
  (issue Manage imports automatically #592).
• Implement the standard methods that were previously stubs:
  signatureHelp, rename / prepareRename, documentHighlight,
  typeDefinition, workspaceSymbol, foldingRange, callHierarchy,
  and completionItem/resolve.
• Cross-language go-to-definition: jumping from a TTCN-3 identifier
  into an ASN.1 module now works.
• References and rename resolve through symbol definitions instead of
  raw name-text matching, eliminating cross-module false positives.
• Switch textDocument/sync from Full to Incremental and splice
  changes into cached buffers per keystroke.
• Flip diagnostics / format / semanticTokens / inlayHint defaults to
  on while keeping the legacy ttcn3.experimental.*.enabled keys as
  opt-out fallbacks.
• Advertise the kinds we actually emit via CodeActionOptions so
  clients can pre-filter via the Only field.
• Cache line-offset lookups in ttcn3/syntax.Root.searchLines with
  correctness + sequential/random benchmarks (issue Add cache for binary search in lines-map #640).

Formatter

• Replace the tabwriter-based CanonicalPrinter with a Wadler-style
  document combinator layer plus a width-aware wrapping formatter,
  configurable via a new [tools.fmt] section in the manifest.
  (Design inspired by Vanadium's PrintDirective vocabulary.)

Lint / analysis

• Mirror the formatter wiring with [tools.lint] for disabling
  individual rules.
• Arena allocator with sync.Pool for hot lint paths.
• Rule lifecycle inspired by Vanadium's Rule / Context.

ASN.1 frontend (internal/asn1)

A pragmatic but full-featured pure-Go frontend covering X.680, X.681,
X.682, and X.683. The cross-module Basket, the
WithSyntaxParser/ObjectSetResolver class driver and the
TTCN-3 lowering pass are direct Go reimplementations of Vanadium's
Asn1ModuleBasket, ClassObjectParser/ClassSetResolver and
Asn1AstTransformer.

• New lexer, AST, and recursive-descent parser with a fully buffered
  token stream for reliable backtracking.
• Semantic resolver with cross-module basket and EXPORTS/IMPORTS
  validation.
• Parameterisation engine with cross-module chained substitution and
  caching.
• WITH SYNTAX / object-set / component-relation class driver.
• Lowering pass that emits parseable TTCN-3 source.
• Hooked into ttcn3/db (ParseFileFull, ASN1Location) and the
  LSP definition handler.
• Legacy asn1.Parse / asn1.ParseFile preserved via an adapter
  layer so existing callers don't break.

Project layer

• Titan project descriptor (.tpd) loader with recursive
  <ReferencedProjects> support, wired into project.Open,
  project.AutomaticRoot, and project.Discover (issue Support Titan project descriptor files #650).

Protocol / runtime

• LSP 3.17 additions: pull-diagnostics, type hierarchy, position
  encoding.
• Schema-driven AST generator under ttcn3/v2/syntax/nodes.
  (Schema layout adopted from Vanadium's src/ast/nodes.yml.)

CI

• Add a 3-OS matrix (ubuntu-latest, macos-latest, windows-latest)
  to the Tests job (issue Fix tests for windows #629).
• Enable core.longpaths=true on Windows before checkout so the
  ETSI conformance suite (paths up to 274 chars) can be cloned.
• Normalise path-sensitive tests across internal/cache,
  internal/fs, internal/lsp, internal/lsp/span, and project
  so they pass on Windows without altering production behaviour.

Docs

• New end-to-end Getting Started walkthrough at
  docs/getting-started.md, linked from README.md (issue Getting started is unfinished #572).
• New THIRD_PARTY_NOTICES.md documenting third-party inspirations
  and licenses.

Linked issues

Closes #592, #629, #640, #650, #572.

Test plan

• go test ./... green on Linux
• go test ./... green on macOS (via CI matrix)
• go test ./... green on Windows (via CI matrix)
• golangci-lint run passes (lint job stays green)
• New ASN.1 fixture + integration suite exercises parser ->
  resolver -> param -> class -> transform end-to-end
• organize_imports action verified against unsorted, sorted,
  single-import, and multi-module inputs
• .tpd loader unit-tested for source/folder/reference resolution
  and recursive ReferencedProjects expansion

[makekryl]

Well, despite the fact that I was not notified about this and noticed this by accident, I, as the Vanadium author, believe I have to say that plagiarizing a very massive, handmade work (which is still in progress and hasn't been abandoned!) with the help of AI agents (I'm not sure if it's possible to port this amount of complex logic otherwise in the in a record-breaking couple of hours, while I spent almost a year on it) - especially without leaving any credit to the original work! - is not a good move (rather, a very bad move) for the future of open-source in general.

If you like the tools I make, you could always contact me with a proposal to work on this.

[rafael2knokia]

Well, despite the fact that I was not notified about this and noticed this by accident, I, as the Vanadium author, believe I have to say that plagiarizing a very massive, handmade work (which is still in progress and hasn't been abandoned!) with the help of AI agents (I'm not sure if it's possible to port this amount of complex logic otherwise in the in a record-breaking couple of hours, while I spent almost a year on it) - especially without leaving any credit to the original work! - is not a good move (rather, a very bad move) for the future of open-source in general.

If you like the tools I make, you could always contact me with a proposal to work on this.

Hi, this was not the intent, mostly an experiment. Your software is pretty complete, wonderful work.
I'll add the appropriate credits and keep this as draft.

[rafael2knokia]

@makekryl You're right, and I'm sorry. Vanadium directly inspired multiple subsystems in this PR — most obviously the ASN.1 Basket / ObjectSetResolver / WithSyntaxParser architecture, the YAML-driven AST schema, the Wadler-style formatter combinators, and the lint Rule lifecycle. The branch name and the first commit title (Close vanadium feature gap…) make that clear internally, but I failed to carry that acknowledgement into the PR description, source files, or a NOTICE entry, and you found out about it by accident. That's not OK and I want to fix it.

I've just pushed 92dbe17 which:

1. Adds a THIRD_PARTY_NOTICES.md crediting Vanadium and you as the author of the original work, with the full BSD-3 copyright/license notice and a list of the borrowed concepts (Asn1ModuleBasket → resolver.Basket, ClassObjectParser / ClassSetResolver → WithSyntaxParser / ObjectSetResolver, Asn1AstTransformer → transform.LowerModule, nodes.yml schema → ttcn3/v2/syntax/nodes, PrintDirective vocabulary → ttcn3/format combinators, Rule lifecycle → ttcn3/lint).
2. Adds attribution in the package-level doc comments of the most clearly derivative files (internal/asn1/resolver, internal/asn1/class, internal/asn1/transform, ttcn3/format/doc.go, ttcn3/v2/syntax/nodes/doc.go, ttcn3/lint/lint.go).
3. Adds an Acknowledgements section to README.md pointing at the notices file.
4. Rewrites the PR description so Vanadium's role is the very first thing a reviewer sees.

I started this as an experiment and apologize any inconvenient.

[makekryl]

My sincere gratitude for refraining from creating diverging same-origin implementations