fix(memory, connection): two data-race/counter defects

1) connection_pool.go: GetConnectionStats read lastRecoveryAttempt (a
multi-word time.Time) without any lock while checkAndRecover wrote it
under recoveryMutex -> data race on live admin API endpoints. Snapshot it
under recoveryMutex (the writer's lock). Regression test
TestGetConnectionStatsRace drives the read path concurrent with the locked
write under -race.

2) cache/memory memory.go: Cache.Get expired path used an unconditional
Delete + atomic decrement, so concurrent Gets on the same expired entry
each decremented entryCount/memoryUsage -> permanent negative drift. Use
LoadAndDelete-with-guard (as evictOldest/evictToFreeMemory already do) so
only the goroutine that actually removes the entry decrements. Regression
test TestMemoryCacheConcurrentGetOnExpiredEntry_CountersConsistent.

cache/memory/memory.go

@@ -181,9 +181,14 @@ func (c *Cache) Get(key string) ([]byte, bool) {
 
 	cacheEntry := entry.(CacheEntry)
 	if cacheEntry.ExpiresAt.Before(time.Now()) {
-		c.entries.Delete(key)
-		atomic.AddInt64(&c.entryCount, -1)
-		atomic.AddInt64(&c.memoryUsage, -cacheEntry.MemorySize)
+		// Guard the counter decrements with LoadAndDelete so concurrent Gets on
+		// the same expired entry don't each decrement entryCount/memoryUsage
+		// (the unconditional Delete here previously double-counted).
+		if removed, ok := c.entries.LoadAndDelete(key); ok {
+			removedEntry := removed.(CacheEntry)
+			atomic.AddInt64(&c.entryCount, -1)
+			atomic.AddInt64(&c.memoryUsage, -removedEntry.MemorySize)
+		}
 		return nil, false
 	}