[FEA] GFQL local output_type + Table select pipeline for conformance

[lmeyerov]

Is your feature request related to a problem? Please describe.
Cypher TCK conformance work needs row-style outputs for RETURN/ORDER BY/LIMIT and basic aggregations. Today, local gfql() always returns a Plottable and relies on ad-hoc Python DataFrame post-processing. Remote gfql already supports output_type=nodes/edges, but local does not, and there is no in-language table pipeline.

Describe the solution you'd like
Add a minimal, vector-friendly Table pipeline and align local gfql() with remote output_type:

• gfql(..., output_type='all'|'nodes'|'edges') for local execution (parity with gfql_remote).
• nodes_table(g) / edges_table(g) conversion ops (graph -> table), optionally filtered by alias name.
• A single table op: select(table=..., where=..., project=..., order_by=..., group_by=..., agg=..., limit=..., distinct=...).
• Optional: graph(edges=table, src=..., dst=..., nodes=..., node_id=...) as table->graph bridge (hypergraph already exists, so keep this low-level).
  All operations should compile to pandas/cuDF vectorized primitives.

Describe alternatives you've considered

• Keep using Python DataFrame post-processing outside GFQL. This works but blocks conformance and makes translation harder to standardize.
• Add only output_type nodes/edges for local gfql() without any Table ops; still leaves RETURN/ORDER/AGG outside GFQL.

Additional context
Conformance status: 718/1615 scenarios translated; expressions bucket (758 scenarios) and row semantics remain 0% covered. A minimal table pipeline would unlock a large fraction of non-expression RETURN/ORDER/LIMIT cases and reduce Python-only post-processing in the TCK harness.