feat(gfql): deep axis payload validation for ring_*_layout calls

[lmeyerov]

Problem

GFQL ring_*_layout call validation currently accepts axis as only list|dict without deep structural checks.

In practice, malformed axis payloads can pass early validation and fail later at render/interaction time, making errors harder for downstream agents (LLMs) to correct.

Current behavior

In graphistry/compute/gfql/call/validation.py:

• ring_categorical_layout.axis: lambda v: v is None or is_list_or_dict(v)
• ring_continuous_layout.axis: lambda v: v is None or is_list_or_dict(v)

No per-row schema checks.

Additional audit notes (2026-04-29)

• We confirmed this gap while validating Louie GraphAgent.gfql prevalidation behavior.
• Structured payloads are prevalidated (ASTCall.from_json(..., validate=True)), so stronger checks here would immediately improve user-facing errors without Louie-specific logic.
• For radial/ring workflows, this is currently the biggest source of "late" / non-actionable failures.

Requested behavior

Add deep validation for ring axis payloads with explicit field-level errors, e.g.:

• Accept list of rows where rows validate known keys/types:
  • numeric coords: r / x / y numeric when present
  • label string when present
  • flags internal / external / space booleans
• For dict axis forms, validate key/value typing and consistency.
• Reject unknown/ill-typed fields with actionable messages.

Why this matters

• Better early failures for GFQL call users.
• Better LLM/tooling correction loops.
• Less ambiguity between backend and frontend axis rendering behavior.

Related

• Louie-side audit report and repro context: graphistrygpt PR flow feat/gfql-radial-layout-proof
• Companion issues:
  • settings-level strict validators: feat(validation): expose strict reusable settings validators #1240
  • call/encoding safelist parity gaps: feat(gfql): close encoding call parity gaps (encode_edge_icon / encode_axis) + strengthen encode validators #1241