perf(gateway): scale request hot paths for 20k streams

Shard and singleflight hot-path caches, batch and prioritize candidate and usage lifecycle persistence, and extend database and pressure-test instrumentation for 20k concurrent streams.

crates/aether-data/runtime/schema/bootstrap/postgres/180_usage_stale_pending_cleanup_index.sql

@@ -0,0 +1,5 @@
+-- Empty databases do not need a concurrent build: the snapshot runs before
+-- traffic starts and executes inside a transaction.
+CREATE INDEX IF NOT EXISTS idx_usage_stale_pending_created_request
+ON public.usage USING btree (created_at, request_id)
+WHERE status IN ('pending', 'streaming');

crates/aether-data/runtime/schema/bootstrap/postgres/manifest.txt

@@ -12,3 +12,4 @@
 150_background_tasks.sql
 160_routing_profiles.sql
 170_usage_legacy_body_ref_cleanup_index.sql
+180_usage_stale_pending_cleanup_index.sql