test(mcp): refresh context baselines after combined schema growth

[bokelley]

Summary

• refresh MCP context-analysis counts after the proposal-refinement schemas landed
• increase the media-buy model-context ceiling from 384 KiB to 388 KiB
• increase the compact product-discovery ceiling from 128 KiB to 132 KiB

These are the exact failures in the post-merge main Build Check: https://github.com/adcontextprotocol/adcp/actions/runs/31902278746. The contributing feature PRs passed independently against earlier base revisions, but their combined schema growth made main red.

Validation

• npm run test:mcp-schema-projection (24/24 after the final baseline adjustment)
• npx vitest run --config server/vitest.config.ts server/tests/unit/product-discovery-schema-parity.test.ts (4/4)
• npm run typecheck
• full pre-commit suite

[aao-secretariat]

Ladon verdict: Approve

Approve — test-only baseline refresh, no blocking findings.

This PR touches only three test files (server/tests/unit/product-discovery-schema-parity.test.ts, tests/mcp-schema-analysis.test.cjs, tests/mcp-schema-projection.test.cjs), bumping hardcoded MCP schema-context counts and two size ceilings (128→132 KiB compact product-discovery, 384→388 KiB media-buy model-context) to absorb schema growth already merged to main. Counts move consistently (+3 across unique_definitions, dictionary $defs, and both shared_dictionary counts from the same new defs), the standalone context_bytes assertion stays pinned to an exact value, and no published protocol surface (static/schemas/source/, normative docs, dist/) is touched — so no changeset is required.

Decision path: no critical/high findings (row 1 n/a); gated_paths=false so row 2 does not fire despite review_decision=REVIEW_REQUIRED; high_risk=false (rows 3/5 n/a); no medium findings in any escalatable category (row 4 n/a); no prior decision (row 6 n/a); no no-auto-approve team match (row 7 n/a); zero medium findings (row 8 n/a). Falls through to row 9 → approve.