test(tooldiscovery): add search ranking benchmark

[aniruddhaadak80]

Summary

Add a repeatable, inventory-backed benchmark for the existing heuristic tool-search ranking without changing production ranking behavior.

Why

Fixes #2996

What changed

• Add 45 hand-labelled natural-language cases spanning the major GitHub toolsets.
• Build the real all-tools inventory through github.NewInventory and validate every label against the current tool name and toolset.
• Measure Recall@1, Recall@3, and MRR@10 across independently rebuilt inventories.
• Add a standard BenchmarkToolSearchCorpus for repeatable latency and allocation measurements.

MCP impact

• No tool or API changes
• Tool schema or behavior changed
• New tool added

This is test-only and leaves the current ranking implementation unchanged.

Prompts tested (tool changes only)

N/A

Security / limits

• No security or limits impact
• Auth / permissions considered
• Data exposure, filtering, or token/size limits considered

Tool renaming

• I am renaming tools as part of this PR (e.g. a part of a consolidation effort)
  • I have added the new tool aliases in deprecated_tool_aliases.go
• I am not renaming tools as part of this PR

Lint & tests

• Linted locally with ./script/lint
• Tested locally with ./script/test

Passed:

• go test -race -count=1 . in pkg/tooldiscovery
• go test -count=1 . in pkg/inventory
• go vet ./...
• go test -run '^$' -bench BenchmarkToolSearchCorpus -benchtime=1x -count=1 -benchmem . in pkg/tooldiscovery
• Pinned golangci-lint v2.9.0 reported 0 issues for pkg/tooldiscovery

The current heuristic baseline is Recall@1 0.844, Recall@3 0.911, and MRR@10 0.885 across 45 cases. The one-shot Windows benchmark measured 2.33 seconds, 48.9 MB, and 101,283 allocations per corpus run; these performance figures are machine-dependent.

The repository-wide go test ./... run passed all packages except pre-existing Windows toolsnap icon-order mismatches in pkg/github.

Docs

• Not needed
• Updated (README / docs / examples)