Consolidate notebook storage: remove runme-fs-workspaces and mirror fs/contents into runme-local-notebooks

[jlewi]

Summary

We currently have two IndexedDB databases for notebook-related state:

• runme-local-notebooks (Drive/local mirror)
• runme-fs-workspaces (filesystem workspace handles/entries)

Per the latest discussion in #55, we should evaluate consolidating onto a single mirror model in runme-local-notebooks and treating local filesystem + contents:// as upstream/remote URI backends.

Reference: #55 (comment)

Motivation

• Reduce storage model complexity and duplication.
• Make notebook mutation flows consistent across local://, fs://, and contents://.
• Simplify sandbox/runtime API design for notebook manipulation (single persistence model + revision semantics).

Proposed Direction

1. Define a single canonical local mirror model in runme-local-notebooks for all notebook docs.
2. Treat filesystem and contents backends as upstream sources identified by remote URIs (e.g. file://..., contents://...).
3. Remove or greatly reduce runme-fs-workspaces responsibilities.
4. Ensure conflict detection/revision semantics remain consistent across backends.

Scope for this issue

• Design + migration plan (schema, URI mapping, sync strategy).
• Implementation changes in stores/resolver/context wiring.
• Backward compatibility + migration from existing fs://workspace/... entries.
• Tests covering open/list/load/save/rename/conflict flows.

Notes

This issue is cleanup/foundation work and is intended to support the sandboxed AppKernel direction in #154.