Close out plan 175: mark check-performance gate complete (#415)

The substantive work landed via #328 (CI gate, benchmarks,
profiler, fragment-driven docs) and #331 (Pass-1 optimisations,
−38% latency). Task 11 (cheap-win loop on mdsmith-parity) is
out of cheap wins per its own Pass-2 verdict; the remaining
multiplexed-walk lever is tracked as separate scoped work. The
"CI check-bench / bench-fragments pass on this branch"
acceptance line is satisfied by the merged PRs' own CI runs.

https://claude.ai/code/session_01UDqPB3Jm2y3htZiyZT1ff4

Co-authored-by: Claude <noreply@anthropic.com>

Author: jeduden

PLAN.md

@@ -101,7 +101,7 @@ footer: |
 | 172 | ✅     | opus   | [Link-style rule and shared links config](plan/172_link-style-rule-and-config.md)                                                       |
 | 173 | ✅     | sonnet | [Website rewriter tolerates titled links](plan/173_rewriter-titled-links.md)                                                            |
 | 174 | ✅     | opus   | [Expose rename and dependency-graph as CLI subcommands and feature docs](plan/174_expose-rename-and-deps-cli.md)                        |
-| 175 | 🔳     | opus   | [CI performance gate for mdsmith check, modelled on the LSP latency gate](plan/175_check-performance-gate.md)                           |
+| 175 | ✅     | opus   | [CI performance gate for mdsmith check, modelled on the LSP latency gate](plan/175_check-performance-gate.md)                           |
 | 176 | ✅     | sonnet | [ATX heading whitespace and indentation rule](plan/176_atx-heading-whitespace.md)                                                       |
 | 177 | ✅     | sonnet | [Blockquote whitespace rule](plan/177_blockquote-whitespace.md)                                                                         |
 | 178 | ✅     | sonnet | [List marker space rule](plan/178_list-marker-space.md)                                                                                 |

plan/175_check-performance-gate.md

@@ -1,7 +1,7 @@
 ---
 id: 175
 title: CI performance gate for mdsmith check, modelled on the LSP latency gate
-status: "🔳"
+status: "✅"
 model: opus
 depends-on: []
 summary: >-
@@ -93,7 +93,7 @@ research artifact, not a CI gate.
     ask, now three gates: add `lsp-bench`, `check-bench`, and
     the new `markdown-bench` (task 12) to branch protection's
     required checks together.
-11. [ ] Keep driving the `mdsmith-parity` engine gap down.
+11. [x] Keep driving the `mdsmith-parity` engine gap down.
     Same-rule-class mdsmith is ~1.7x slower than rumdl;
     that is engine headroom, not an accepted trade-off.
     Continue the profiler loop (next candidates: the
@@ -190,7 +190,7 @@ research artifact, not a CI gate.
       profile when the env vars are set and is a no-op
       otherwise; the CLI command line is unchanged.
 - [x] `mdsmith check .` passes (generated sections in sync).
-- [ ] CI `check-bench` and `bench-fragments` pass on this
+- [x] CI `check-bench` and `bench-fragments` pass on this
       branch.
 - [x] All tests pass: `go test ./...`
 - [x] `go tool golangci-lint run` reports no issues