chore(#649): the required check's name stops being a prose string

[gsdali]

Part of #649 — the rename that must land before the ruleset is applied.

Why

gate-scripts is becoming a required status check on refactor/**. The context such a rule matches is the check-run name, which GitHub takes from the job's name: when it has one and from the job id otherwise.

Today that name is the prose string gate scripts (static, no build). It reads like a comment. If anyone rewords it for clarity, the ruleset keeps pointing at the old string, the check quietly stops being required, and the PR UI looks identical to a working gate.

That is the #618 failure class — a detector reporting all clear because it is blind, indistinguishable from one reporting all clear because the tree is clean — reproduced one level up, in the enforcement rather than the detector. Three gate scripts on this branch already shipped confidently wrong; this makes the thing that enforces them harder to disarm by accident.

Dropping the key makes the required context the job id, which nobody edits for readability. Cost: a less readable label in the checks list.

Ordering

This must merge before the ruleset is created. Applying a rule that requires gate-scripts while the job still publishes gate scripts (static, no build) would block every PR into refactor/** with an unclearable "Expected, waiting for status to be reported" — including this one.

Also recorded in CLAUDE.md

• The requirement itself, and why the name: key is deliberately absent.
• It is not required on main, and must not be until ci.yml carrying this job lands there. For a pull_request the workflow is read from the merge ref, so once the base has the job every PR gets it regardless of how stale the head is — but until then a required check would never report.
• build-and-test is deliberately not required on refactor/** (0-for-21 under CI never validates a kernel-patch PR against the patched kernel it claims to fix #585). It is ~9/10 green on main, so the requirable set inverts per branch. CI never validates a kernel-patch PR against the patched kernel it claims to fix #585 is confined to this branch and its descendants, not repo-wide — a correction to how it has been described.

Verification

YAML parses; jobs unchanged at gate-scripts, build-and-test, ios-simulator-build; gate-scripts now has no name: key. All four gates pass locally (exit 0). The check-run name this PR's own CI run publishes is the confirmation that matters, and it is read from the API rather than the PR UI before the ruleset goes anywhere near the repo.

build-and-test is expected red (#585, signature Issue570HealingApproxTests (fitted.uDegree → 1) > 1).