Wire the repo-factory phase gate into CI

[MongLong0214]

Closes #103.

`phase-gate.py` (repo-factory skill's Phase 4/5/6 completion gate, used to
decide when this project and gitseed have finished a phase) lives at
`~/.claude/skills/repo-factory/scripts/`, a path no CI runner can reach.
`docs/ROADMAP-TO-DONE.md` (Goal 3) recorded this as outstanding. This adds
`.github/workflows/phase-gate.yml`, which fetches the script from a new
public mirror (`MongLong0214/repo-factory-gate`) pinned by commit SHA +
sha256 and runs it with `GH_TOKEN` set, so the gate that decides phase
completion can actually run -- and fail the build -- on the runner. Also
updates the ROADMAP row to record the resolution.

• Phase 4 / 5: `workflow_dispatch` (on-demand — Phase 5's own condition is
  false for most of an active wave by construction: Phase 4: design and run M4 after M3 was voided #52 is open right now,
  non-Backlog, correctly)
• Phase 6: also automatic on every `vX.Y.Z` tag push (its conditions stay
  met once met)

Options ruled out (also recorded as commit trailers): vendoring a copy into
this repo and gitseed (a second copy free to drift from CommitLore's own
copy of the same trust logic — the defect class #89/#90 closed); moving the
gate into this repo and having the skill reference it (inverts the
dependency).

Not merging to `dev` myself per instructions — this is for review.

[github-actions]

CommitLore — record lint

Trailers: clean — 2 commits in origin/dev..aa89670801d20b136a604f51e226efd96f093d23
Active constraints: 1 limits · 5 ruled-out · 3 warnings — from 4 records over 2 changed paths

Active constraints for the paths this PR touches

Limits (1)

• - c993bda — the mirror repo is published-to, not edited directly -- patch the skill, then re-publish

Ruled out (5)

• - c993bda — vendor phase-gate.py into this repo and gitseed | second copy free to drift from CommitLore's own trust logic -- the defect class parse and context disagree about how many records a message contains #89/Shape verdict depends on repository presence, contradicting SPEC §6.1's own class definition #90 closed
• - c993bda — move the gate into this repo, have the skill reference it | inverts the dependency -- a project should not own the factory's gate
• - c993bda — gate every dev push on Phase 4/5 | Phase 5's own condition (Phase 4: design and run M4 after M3 was voided #52 open, non-Backlog) is false mid-wave by construction; trains reviewers to ignore CI
• r-3b57e2 30f2d5f — converting the three translated READMEs for consistency | they are the product, not the record, and two checks exist specifically to keep them
• r-5a83e9 1c683df — requiring a positive benchmark for release | a tool that answers honestly ships whether or not the effect turns out to be large, and pretending otherwise is what the withdrawal of the published numbers was about

Warnings (3)

• r-3b57e2 30f2d5f (claim) — bench/PREREGISTRATION.md is append-only and was translated in place. Its section numbering and order are unchanged, but a translation is still an edit to a file whose whole discipline is that it is not edited. Recorded here rather than left to be noticed
• r-5a83e9 1c683df (claim) — the claim that gitseed's rejected alternatives will produce a higher control base rate is a prediction, not a measurement. The qualification round exists to test it, and it may say no
• r-1a63f5 2bb4993 (claim) — "CI is green" was said five times today against a red CI, including in the commit that introduced the rule saying to check CI before saying it. The rule is in docs/RELEASE-GATE.md §5 and it was not followed by its own author. This commit is not claiming CI is green; that claim comes after the run reports

git log --follow accepts exactly one pathspec, so renames are not followed for 2 paths; query one path at a time to follow its rename chain

_{Trailer violations fail this check. Active constraints are informational — they are what the repository already decided, not a verdict on this PR.}

[MongLong0214]

Closing: repo-factory's scope is idea-to-initial-repo only (concretize an idea, then set up/configure the repo until it's ready for work to begin). Wave execution and shipping are not its job, and the skill must not couple to any project's CI. This PR was the wrong direction — reverting.