What does this PR do?

Slice 1 of 5 splitting #1706, in the order requested in review. This one stands alone and depends on nothing else in the series.

tree-sitter-ruby wraps class C < Base in a superclass node whose child is a constant (or a scope_resolution for A::B). collect_bases_from_field matched neither, so the superclass fell through to the raw-text fallback and base_classes stored "< Base" — operator included. A base spelled that way never resolves, so Ruby subclasses produced zero INHERITS edges and the Rails chain (User < ApplicationRecord < ActiveRecord::Base) never linked up.

The fix adds both node kinds to the allow-list that already carries Python's identifier / attribute.

Reproduce-first. The new inherit_ruby rows in tests/test_extraction_inheritance.c fail 5/5 without the one-line change, each on the "<" substring that is the bug's signature:

FAIL [Dog] expected base "Animal" not found in base_classes
FAIL [Dog] bad string "<" found inside a base_classes entry
... 5 / 5 cases above failed

Incidental observation, not changed here: inherit_python, inherit_typescript, inherit_tsx, inherit_php and inherit_kotlin all pass now, so the RED: markers on them in SUITE(extraction_inheritance) are stale. Left alone to keep this PR to one thing; happy to file it separately.

Where this sits in the series

Slices 1, 2 and 3 are independent: I verified the resolver in slice 3 passes all its rows with neither extraction fix applied, so they are only stacked where a later slice genuinely needs an earlier one, per the review.

CI baseline

main at 4d7d9f1 is red for reasons unrelated to this change: #1181 gave list_projects pagination parameters, retiring the last empty-properties schema that #1359's guards used as their zero-argument example. cli_zero_argument_tool_never_reads_stdin_issue1359 and cli_stdin_args_gate_tracks_tool_schema_issue1359 fail on every leg. #1704 already fixes it. Until that merges, ci-ok will be red here for that reason.

Refs #1701

Checklist

• Every commit is signed off (git commit -s) — required, CI rejects
  unsigned commits (DCO, see CONTRIBUTING.md)
• Tests pass locally (scripts/test.sh) — full canonical leg run on the five slices merged together: 7542 passed, 32 failed, 5 skipped. All 32 failures are in cli and reproduce identically on unmodified origin/main (2 are the fix: unbreak main — stdin-gate #1359 guards + Windows search_code UTF-8 pipe #1704 breakage above; 30 are this machine's real agent configs interfering with the CLI install tests). No new failure is introduced by this series.
• Lint passes (make -f Makefile.cbm lint-ci)
• New behavior is covered by a test (reproduce-first for bug fixes)

Assisted by OpenCode using anthropic/claude-opus-5.