fix: Add sync health checks to all query methods for consistent repair triggering

Two fixes:

1. Sync query methods (query_as_of_bsatn, query_as_of_pk_bsatn,
   query_time_series_bsatn, query_time_series_pk_bsatn) now detect
   missing/repairing segments and trigger repair via fire-and-forget
   spawns on the temporal runtime. Previously these methods had no
   health checks, so enabling temporal on existing databases would
   return "No data found" for AS OF queries via SQL without ever
   triggering repair. ALL BETWEEN worked because its async path
   had full health checks.

2. Grid alignment checks in build_ops.rs moved inside per-transaction
   loops for future-proofing when callers start batching multiple
   transactions per commit (upstream PR clockworklabs#4404 removed the n=1 limit).

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: Ludv1gL