Fix status:* label hygiene in the ZenDev runbooks

[drevendev]

Closes #20

Achieved outcome

Status-label hygiene now holds across the AUTHOR/ACCEPTOR loop: setting a
status:* label always replaces whichever one was already there, and closing an
Issue removes its status:* label. The seven closed Issues that had already
accumulated a stale label (#19, #17, #14, #10, #8, #6, #4) are cleaned up directly.

Tested revision

d86b979bfd27cffd7b1648aae69f78f35918cc8e

Changed artifacts

• docs/zendev/AUTHOR_RUNBOOK.md — section 3 (claim) now says setting
  status:in-progress replaces any existing status:* label; section 7 (hand off)
  says status:needs-review replaces status:in-progress; section 8 (blocked) says
  status:blocked replaces any existing status:* label.
• docs/zendev/ACCEPTOR_RUNBOOK.md — section 4 (verdict) now says that once an
  Issue is confirmed closed, every status:* label it still carries is removed, and
  names the fallback when the merge does not close the Issue automatically: close it
  yourself, then remove the label(s).
• Forge-side only (not part of the diff): removed all status:* labels from closed
  Issues IMPLEMENTATION_STATUS goes stale the moment a pull request merges #19, Lock a hash-verified seeded legacy baseline (REQ-MIGRATION-002) #17, Give canonical TypeScript work a required check #14, Mirror only the allowlisted specification package #10, Record what each autonomous run consumes, privately #8, spec-sync fails: RCLONE_VERSION is consumed by rclone as a flag #6, Establish the autonomous development loop for this repository #4, per the Issue's correction comment.

Acceptance criteria

• The acceptor runbook states that closing an Issue includes removing its
  status:* label — docs/zendev/ACCEPTOR_RUNBOOK.md section 4.
• Issue Lock a hash-verified seeded legacy baseline (REQ-MIGRATION-002) #17 carries no status:* label — verified with
  gh issue list --state all --json number,labels.
• The rule names what to do when the Issue does not close automatically —
  ACCEPTOR_RUNBOOK.md section 4, added paragraph.
• (from the correction comment) No open Issue carries more than one status:*
  label, and no closed Issue carries any — verified with
  gh issue list --state all --json number,labels: Closed Issues keep an active status label #20 (open) carries only
  status:in-progress; IMPLEMENTATION_STATUS goes stale the moment a pull request merges #19, Lock a hash-verified seeded legacy baseline (REQ-MIGRATION-002) #17, Give canonical TypeScript work a required check #14, Mirror only the allowlisted specification package #10, Record what each autonomous run consumes, privately #8, spec-sync fails: RCLONE_VERSION is consumed by rclone as a flag #6, Establish the autonomous development loop for this repository #4 (closed) carry none.
• (from the correction comment) The author runbook states that setting a
  status:* label replaces any existing one — AUTHOR_RUNBOOK.md sections 3, 7, 8.

Checks

Check	Outcome	Evidence
npm ci	passed	39 packages installed, 0 vulnerabilities
npm run typecheck	passed	tsc --noEmit clean
npm test	passed	vitest: 1 file, 2 tests passed
npm run build	passed	vite build succeeded
dotnet restore	passed	both projects restored
dotnet build --configuration Release	passed	0 warnings, 0 errors
dotnet test --configuration Release	passed	44/44 passed
python -m unittest discover -s scripts/tests	passed	20/20 passed
python scripts/policy_guard.py --base origin/master	passed	"passed over 2 changed file(s)" — docs/zendev/** only, no product paths

All checks run locally against the tested revision above; no check is not_run or
unavailable.

Not checked

Nothing deliberately skipped. This is a docs-only change to two runbook files; there
is no product behavior to regress, so no new automated test was added — the change
is process guidance, verified by re-reading the edited sections and by the label
state shown above.

Assumptions and unknowns

• Assumption: the Issue's acceptance criteria (as narrowed by the correction
  comment) are the full scope of this unit of work; the correction comment states
  "Nothing else about the Issue changes," which I took as authoritative over the
  original body wherever the two differ.
• Assumption: "the label catalog allows at most one status:*" (stated in the
  correction comment) is an existing catalog rule this PR documents in the runbooks,
  not a new rule being introduced — gh label list shows the six status:* labels
  as a single-select-by-convention axis, consistent with AGENTS.md's Issue
  standard.
• The forge-side label removals on IMPLEMENTATION_STATUS goes stale the moment a pull request merges #19, Lock a hash-verified seeded legacy baseline (REQ-MIGRATION-002) #17, Give canonical TypeScript work a required check #14, Mirror only the allowlisted specification package #10, Record what each autonomous run consumes, privately #8, spec-sync fails: RCLONE_VERSION is consumed by rclone as a flag #6, Establish the autonomous development loop for this repository #4 are not visible in
  this diff (labels are not repository files); they're recorded here and on the
  Issue as the durable evidence.

Highest-risk area for review

The three edited points in AUTHOR_RUNBOOK.md (claim, hand-off, blocked) are the
only enforcement mechanism for "at most one status:* label" going forward — it's a
runbook instruction, not a workflow-enforced guard, matching the Issue's explicit
non-goal ("No automation of label hygiene in a workflow"). If a future AUTHOR run
skips the replace step, the label can drift again; that's an accepted, named
trade-off, not an oversight.

Remaining gate

None. No decision, authority, or follow-up is required to accept this change.