Correlate registration policy buckets by id instead of key when editing

[nbudin]

Fixes #11870

RegistrationPolicy.build_from_hash now keeps a bucket's id (only the policy-level id is still stripped, to avoid aliasing a real persisted row). RegistrationPolicy#sync_buckets_from_hash! matches buckets primarily by id, falling back to key-matching when the incoming hash doesn't carry an id — so callers that never round-trip an id (e.g. hand-built test hashes, import_convention_data_service) keep working exactly as before.

This means an edited-but-existing bucket now keeps its real row/identity even if its key changes, instead of being destroyed and recreated. The production editing flow (update_event/update_event_proposal via form_response_attrs_json) already round-trips bucket id today via Apollo cache data, so this take effect without any frontend changes.

Deliberately out of scope (per the issue, deferred to follow-up work): converting SignupBucketFinder, RegistrationPolicyBucket#signup_definitely_occupies_slot_in_bucket?, RegistrationPolicy#bucket_with_key, EventChangeRegistrationPolicyService's key-based removed-bucket/mapping logic, or the bucket_key/requested_bucket_key computed readers to work by id — and removing RegistrationPolicyBucket#key itself.

🤖 Generated with Claude Code

[github-actions]

Code Coverage Report: Only Changed Files listed

Package	Base Coverage	New Coverage	Difference
app/javascript/setupI18Next.ts	🟠 65.52%	🟠 72.41%	🟢 6.89%
app/models/registration_policy.rb	🟢 98.59%	🟢 98.68%	🟢 0.09%
Overall Coverage	🟢 55.99%	🟢 56.03%	🟢 0.04%

Minimum allowed coverage is 0%, this run produced 56.03%