fix(quota): honor explicit blocker precedence

[huangruiteng]

Summary

• let an explicit, reasoned blocker claimed by the current agent own the vision-wait read model ahead of an unrelated deferred successor
• treat the authoritative current-agent blocker as a no-replan frontier rule after a blocker delta ACK clears an existing obligation
• keep the exact deferred successor visible in the agent-scope diagnostic projection and add precedence regressions

Why

PR #2530 made blocker identity and reason visible, but the live quality-qualification lane still selected an older exact deferred successor first and immediately re-derived a stalled-monitor replan. That contradicted the documented blocker transition: the repair existed in state but could not own the frontier.

Validation

• 449 passed in tests/control_plane
• focused blocker/vision/rule tests: 45 passed
• Ruff and Python compile checks passed
• loopx check: 0 errors, 0 warnings, public boundary clean
• exact committed diff loopx canary premerge --from-git-diff --git-diff-base origin/main --tier standard: 17/17 checks passed, no manual holds, self-merge allowed
• live registry shadow with this head no longer emits autonomous_replan_required; the repair todo becomes the selected normal-run advancement while it is open

Boundaries

No private state, raw logs, credentials, local paths, benchmark semantics, permissions, or production actions are included.

[huangruiteng]

Maintainer self-review:

• Changed surfaces: goal vision-wait precedence and goal-frontier replan rule selection, plus focused control-plane regressions.
• Behavioral judgment: an explicit typed blocker with a non-empty reason and an exact current-agent claim is a stronger statement of the lane's present frontier than an unrelated deferred successor. It now owns the wait projection. The deferred successor remains visible in agent_scope_frontier.blocked_successor_wait_candidates.
• Safety judgment: missing-reason and other-agent blockers remain non-authoritative. Strict vision gaps, required checkpoints, permissions, and production behavior are unchanged.
• Validation: 449 control-plane tests passed; focused blocker/vision/rule tests passed 45/45; Ruff, Python compile, and git diff --check passed; loopx check reported 0 errors and 0 warnings with a clean public boundary; exact committed diff premerge passed all 17 selected checks.
• Failures/skips/manual holds: none.
• Coverage rationale: the rule-table test proves blocker precedence over a stalled monitor, the integration test proves blocker precedence over an exact deferred successor while retaining successor diagnostics, and the complete control-plane suite covers surrounding quota/status behavior.

The diff is narrow, public-safe, and the risk gate reports self_merge_allowed=true. I consider it ready for self-merge after the required GitHub check passes.