bench: add multirepo benchmark workspace coverage

[autholykos]

Summary

Add benchmark coverage for multirepo workspaces so we can measure the cost of newer cross-repo indexing and analysis behavior, not just single-repo fixture flows.

Why this matters

• multirepo support is now part of the product contract
• current benchmarks do not exercise repo-root binding, repo-labeled output, or cross-repo source sets
• performance characteristics may change once multiple roots are indexed together

Proposed shape

• create a benchmark profile that uses:
  • the Pituitary repo as the primary workspace
  • one sibling repo such as ccd-guide as a secondary repo root
• configure sources with explicit workspace.repo_id, [[workspace.repos]], and repo = "..." bindings
• benchmark at least:
  • rebuild
  • search-specs
  • check-doc-drift
  • analyze-impact

Scope guardrails

• keep this benchmark opt-in
• do not add multirepo requirements to default CI
• keep benchmark assertions focused on performance and successful execution, not correctness duplication

Acceptance criteria

• a documented multirepo benchmark profile exists
• benchmark output includes repo-aware results without ambiguity
• at least one rebuild path and one analysis path run against a logical multirepo workspace
• docs explain how this differs from single-repo benchmarks