perf(server): High-performance backend optimizations

[arvarik]

Summary of Accomplishments

This Pull Request implements the approved backend architectural optimizations for contrack to eliminate critical N+1 database bottlenecks and enhance client request resiliency.

1. Connection Abort Propagation (Security & Cost Control)

• Bound an AbortController to Express req.on("close") in /api/search/semantic.
• Propagated the signal down to the AI reranking service and adapter facades, allowing immediate socket tear-down during early client exits and saving redundant LLM token spending.

2. O(1) Batch Normalization (Dedupe Performance)

• Optimized dedupe scanning in normalizeContacts to batch-load tags and interests at start, mapping them via in-memory O(1) lookups.
• Latency reduced from ~3,000ms down to ~20ms for 1,082 records (a ~150x speedup).

3. High-Performance Contact Bulk Hydration

• Rewrote hydrateMany in contactRepository.ts to leverage query-parameter safe chunked SQL queries (chunk size of 500) and O(N) map-joins.
• Keeps individual microsecond cached statements for collections <= 3 contacts to maintain maximum efficiency.
• Strips contactId fields from child records to guarantee 100% downstream API payload shape parity.
• Yields a ~6x hydration speedup (from 68ms down to 11ms for 500 records).

4. Validation

• Run integration benchmark validating perfect structural equivalence between the sequential and bulk hydration pipelines.
• Fully verified via test harness: 180/180 tests passing successfully.