perf: stale-while-revalidate caching, bigger executor, cheap DB health probe

Third on-box run: p50s healthy everywhere but tails at 60s — when a TTL
expired under load, every concurrent miss parked an executor thread on
the single-flight lock, exhausting the ~12-thread pool and timing out
unrelated endpoints. All cached endpoints (latest, HII, stats, health)
now use _cached_swr: fresh -> inline; expired-but-present -> the stale
value is returned immediately and ONE background task refreshes; only a
cold key (first request since startup) waits. Plus: dedicated
ThreadPoolExecutor (EXECUTOR_THREADS, default 48) replaces the cpu+4
default, and DatabaseHealthCheck no longer re-runs the CREATE TABLE DDL
suite on every probe (connect only when no live engine).

src/config.py

@@ -94,6 +94,11 @@ class Config:
     # TTL for /health check results (includes an external RID-API probe)
     HEALTH_CACHE_TTL_SECONDS = int(os.getenv("HEALTH_CACHE_TTL_SECONDS", "30"))
 
+    # Thread-pool size for blocking work in the web process (DB queries,
+    # inference, health probes). Waiting threads are cheap; starving the pool
+    # stalls every endpoint that needs a thread.
+    EXECUTOR_THREADS = int(os.getenv("EXECUTOR_THREADS", "48"))
+
     # Web server worker processes. Above 1, uvicorn forks workers and a
     # localhost lock port elects a single background-collection leader.
     WEB_WORKERS = int(os.getenv("WEB_WORKERS", "2"))