Flip GITPUSH_RETIRED on by default: the cross-host cutover the stub rehearses

OPS-177 delivered the stub and proved it (bfde099, suite 13/13, four attributive forced-red mutants). This row is the cutover itself, which the stub deliberately does NOT perform: GITPUSH_RETIRED defaults to 0 and nothing flips it. Separate row per pmmaster-evolutiva-cc 2026-10-05: distinct, larger, cross-host scope, not part of "did the stub work". Three legs, none of which is a code change inside sh.git: 1. ~/.claude/git-workflow.md on venus, whey and lezama still names `gitpush.sh "what changed"` as THE command. Per-host, outside this checkout, nw lane. ORDER MATTERS: flipping the default before the doc changes breaks every agent's push mid-session, including the agents who would push the fix. The doc must point at gitpush-chooser.md FIRST. 2. Five executing callers in this checkout: ship.sh:114, deploy.sh:191, gitpush-locked.sh:66, channel-rule-install.sh:59, fixcodexgitpush.sh:7. Four are mechanical substitutions. deploy.sh:191 is NOT: it runs `gitpush.sh --patch`, which bumps the file named by .bump, whereas gitpushhtml.sh bumps through sitebump.sh and the <span class="version"> / HTML_PATH= contract. Which is correct depends on the consuming project and is not knowable from sh.git - that leg needs the project owner, not a find-and-replace. 3. Six suites take the monolith as their SUBJECT, ~100 checks: gitpush-bumporder, gitpush-onlystage, gitpush-scoped-anyrepo, gitpush-prestage-abort, gitpush-gate-test, gitpush-core.test. They retire with it or their cases move into the lane suites. A suite deleted with the script it describes takes its coverage with it, and some of those cases describe gitpush-core.sh behaviour that OUTLIVES the monolith. DONE means: the doc changed on three hosts, no executing caller in sh.git invokes gitpush.sh, the six suites resolved either way with surviving cases rehomed, and only THEN the default flips. Deleting gitpush.sh is a further step, not in this row.