fix(mcp): return semantic-only graph results as primary

[dpersek]

What does this PR do?

Fixes #915.

search_graph now returns vector-ranked semantic hits in the primary results array for semantic-only semantic_query calls, instead of returning broad structural results that can make vector-less nodes appear to win the ranking.

Combined structural + semantic calls still expose semantic_results, and non-array semantic_query values still return the existing validation error. The vector search path now also reports a full semantic candidate count so total and has_more stay meaningful with limit/offset pagination.

Local full-suite caveat: make -f Makefile.cbm test is not green in this environment, but the failing graph-contract/probe slice reproduces on a clean upstream/main worktree, so it is not introduced by this patch. Focused MCP proof and lint-ci are green.

Checklist

• Every commit is signed off (git commit -s) — required, CI rejects
  unsigned commits (DCO, see CONTRIBUTING.md)
• Tests pass locally (make -f Makefile.cbm test)
• Lint passes (make -f Makefile.cbm lint-ci)
• New behavior is covered by a test (reproduce-first for bug fixes)

[DeusData]

Thanks for the semantic_query fix for #915. Triage: high-priority query correctness bug.

Review should verify semantic-only calls put vector-ranked hits in the primary results array, combined structural+semantic behavior stays compatible, and total/has_more report the semantic candidate set rather than the broad structural fallback.