docs: align implementation guides with registry and published results

[tappe9]

Summary

Closes #71.

• Centralize registration, registry-wide CI coverage, official cohort membership, verified publication, and historical introduction/release boundaries in docs/IMPLEMENTATIONS.md.
• Replace stale Flask, Axum, and Express cohort claims with canonical links. Remove Axum's fixed-size CI-matrix claim and Java's ordinal registration count while retaining Java's registered/CI-covered but not officially measured status.
• Distinguish current Flask/Gunicorn from the immutable September 15, 2026 Waitress 3.0.2 report, preserving measured source and version provenance.
• Add five standard-library regression tests to the existing test_benchmark_*.py discovery pattern. Cohort members/order and active status derive from the registry; tests check important boundaries and canonical links/anchors without exact-paragraph snapshots or duplicated implementation counts.

Scope

Baseline: freshly fetched main eba127e65761ced4d0fef06a6c52cc2dc1fdb6ab. Issue #71, comments, open issues/PRs, successful main CI/Pages, AGENTS.md, registry, report metadata and related public guides were inspected before implementation. There was no competing open PR.

Exactly five implementation guides and one test file change. The initial implementation was uploaded together in one commit; a second, formatting-only commit applies the exact pinned-Ruff patch reported by CI. Both are in this one branch/PR and will be squashed.

Root-tree comparison confirms that applications, dependencies, database, registry/measurement code, Compose, workflows, both READMEs, site and all results/** remain identical to baseline. Historical four-stack/schema examples and v0.1.0 boundaries are preserved. Local plans/investigation notes are not committed.

No official benchmark, manual performance experiment, new result publication, release, permission or protection change was requested or performed. Existing normal-CI smoke and Axum diagnostic are unchanged. #52 stays deferred; #59 is reserved for the next separately authorized official producer-to-Pages verification. No work on #72 or later issues is included.

TDD and local evidence

GitHub DNS and the full repository checkout were unavailable in the execution container, as were Docker, actionlint and Ruff. Local tests therefore used an isolated snapshot of fetched source bytes verified against GitHub blob SHAs, not a claimed full checkout or substituted implementation files.

• RED: final regression assertions against original verified documents produced seven expected assertion/subtest failures, with no test errors.
• GREEN: all five documentation regressions passed normally and with python -O.
• Negative checks rejected a wrong active-cohort role and a missing linked heading; exact document bytes were restored and both green runs repeated.
• Python 3.10 syntax parsing and git diff --check passed locally. Actual Python 3.10 execution is covered by CI, not claimed as local execution.
• All six final file blobs match the locally tested bytes.

CI and formatting correction

Initial run 36328802711 failed the pinned Ruff formatting gate. Python 3.10's 193 benchmark tests, 38 contract tests, actionlint and 21 workflow tests had passed, but later shared gates were skipped and that run was not accepted for merge.

Its formatting artifact 10934499130 was downloaded and its SHA-256 verified against the job log. The patch only wraps one assertion call; Python ASTs before/after are identical. The corrected test blob is 4ea6e1dc31c24f7d90f0d0500dbaf2d057afc387.

Latest head: 5eb7084dd4fb01324c4f3312c188991a0edaa6ce.
Normal CI: 36329214231, attempt 1. No manual rerun was requested. Its shared gates, all registered implementation jobs, contracts and Axum diagnostic have passed; smoke and the final required aggregate must also succeed before merge. The final pre-merge audit will record completion against this exact head.

Full local Make/native gates were unavailable. CI is the separate full-environment evidence for registry/Compose, benchmark tests (Python 3.10 and normal/optimized pinned Python), site, DB, Ruff, README generation, source checks, all implementation/contracts and non-publishing checks. make test-workflows was not invoked locally; normal CI runs its workflow-test and actionlint commands directly.

Merge and post-merge boundary

The maintainer explicitly authorized Issue #71 implementation, this PR, squash merge and ordinary post-merge CI/Pages verification. Refresh main, PR head, checks, comments and reviews immediately before merging and pass the verified head SHA to the squash request. Verify automatic main CI and its existing Pages route afterward. Deployment/artifact verification must be distinguished from a direct check of publicly served bytes.

[tappe9]

Final pre-merge self-review completed for head 5eb7084dd4fb01324c4f3312c188991a0edaa6ce.

• Normal CI run 36329214231, attempt 1, is completed/success. All nine implementation jobs, shared checks, registry/Compose plan, non-publishing smoke, Axum diagnostic and cleanup passed. The exact-head required check 108649512160 completed successfully.
• Shared checks include Python 3.10 compatibility, pinned Ruff formatting/lint, workflow tests/actionlint, README generation/source checks, make test-registry, make test-site, make test-db, and normal/optimized benchmark tests. This is CI evidence, not a claim that unavailable local native/Make gates were executed.
• Re-reviewed the final complete diff: only five guides and the new regression file change. Every other root blob/tree matches the freshly fetched baseline, including the complete results tree b06ab3fdc80b21d8d1d6e044a79e1fc90ddbc771. Historical result provenance and frozen cohort/schema examples remain unchanged.
• Immediately refreshed main (eba127e65761ced4d0fef06a6c52cc2dc1fdb6ab), PR head/base/mergeability, CI/required, conversation comments, reviews and inline threads. No additional review findings or conflicting changes were present; the PR is open, non-draft and mergeable.

Proceeding with the maintainer-authorized squash merge using the verified head as the expected SHA. Post-merge main CI and its automatic Pages deployment will be checked separately. No official benchmark or result publication has been dispatched.