feat(semantics): enforce bounded producer ownership

[songoow]

Refs #4447
Related: #4459

Summary

This PR implements the next executable stage of the semantic-vocabulary-convergence-v0 RFC: the M0.5 scope and bounded producer foundation. It makes the producer/consumer distinction operational without claiming whole-program data-flow completeness.

Implemented stage

• adds explicit bounded-context scope declarations for multi-value forks;
• keeps raw inventory fork counts visible while accounting declared semantic forks separately;
• changes legacy should-run retirement measurement to standalone identifier tokens;
• adds a canonical EffectiveAction enum owner for the finite decision-slot value domain while preserving the string wire field;
• adds producer-site metadata and a bounded producer write scan;
• rejects unregistered producer literals even when no consumer compares them;
• validates producer symbols, compatibility-only metadata, and that discovered producer modules are declared;
• keeps dynamic producer paths explicitly outside the proof boundary;
• adds negative tests proving comparisons are not treated as writes and unknown writes fail closed;
• regenerates the semantic inventory after adding the owner carrier;
• keeps the English and Chinese RFC mirrors aligned with the M0.5a/M0.5b contract.

Staged RFC plan

• M0.5: scope declarations, producer closedness/liveness, and identifier-based retirement evidence — this PR.
• M1: replace decision-slot literals with the owner enum at producers and consumers, and split replay/frontier slots.
• M2: publish route/disposition mappings and the decision table through generated cross-runtime bindings.
• M3: retire legacy should-run fields one field per schema-reduction PR after reader proof.
• M4: reduce Python/TypeScript twins with replacement-first migration evidence.

The later stages remain explicitly unproved until their runtime call sites and compatibility evidence are implemented.

Review and validation

Reviewed with the project-owned LoopX premerge flow, using PR #4459's exact-head/evidence pattern. The review found and fixed an assignment-regex bug that could classify == comparisons as writes; validation was rerun after the fix and after rebasing onto the latest upstream/main.

• python3.11 -B examples/semantic-vocabulary-drift-smoke.py — passed
• python3.11 -m pytest -q tests/architecture/test_semantic_inventory.py tests/architecture/test_semantic_vocabulary_drift.py — 24 passed
• python3.11 -m ruff check ... — passed
• git diff --check upstream/main...HEAD — passed
• DCO trailer check for the PR range — passed
• loopx canary premerge --from-git-diff --git-diff-base upstream/main --format json — 14 selected checks passed, 0 failures, 0 manual holds
• loopx check public-boundary validation — passed

[songoow]

Superseded by #4480, as documented in its current description. #4480 carries the M0.5 scope/producer/retirement work with the follow-up fixes for exact producer attribution, liveness evidence and input domains; its refreshed head 0ee2831 includes main edcddc7 and passes all 72 semantic inventory/production/drift tests. Closing this earlier proposal avoids maintaining two competing implementations and repeatedly regenerating both snapshots. The branch and discussion are preserved; no source branch was deleted and no PR was merged.