stream: move addrConnStream.mu to cache line 2 to eliminate false sharing

sentLast, receivedFirstMsg, and decompressorSet are all written without
holding mu (they are written in CloseSend and recvMsg during normal stream
operation). After the tail-grouping commit they landed at offset 232–234
(cache line 3: 192–255), the same cache line as mu (offset 224). Concurrent
writes to these fields invalidate the cache line before every mu.Lock(),
adding up to ~18× overhead under concurrent load.

Moving mu to immediately before parser places it at offset 184 (cache line 2:
128–191). The tail bools remain at offset 232+ (cache line 3), so stressor
writes no longer affect mu.Lock() latency.

Benchmark before this commit (sentLast stressor):
  stressors=0   3.9 ns
  stressors=1  19.1 ns  (4.9×)
  stressors=4  69.1 ns (17.7×)

Benchmark after (different cache lines):
  stressors=0  3.7 ns
  stressors=1  3.7 ns  (~1×)
  stressors=4  3.9 ns  (~1×)

No size change: addrConnStream remains 240 B.

Fixes #9348

Author: gidotencate

stream.go

@@ -1502,12 +1502,14 @@ type addrConnStream struct {
 	codec            baseCodec
 	sendCompressorV0 Compressor
 	sendCompressorV1 encoding.Compressor
-	decompressorV0   Decompressor
-	decompressorV1   encoding.Compressor
-	parser           parser
+	decompressorV0 Decompressor
+	decompressorV1 encoding.Compressor
 
 	// mu guards finished and is held for the entire finish method.
-	mu sync.Mutex
+	// Placed here so it falls on cache line 2 (offsets 128–191), keeping it
+	// off the cache line shared with the tail bool fields (offset 232+).
+	mu     sync.Mutex
+	parser parser
 
 	// Bool fields are grouped at the tail to eliminate the alignment padding
 	// that would otherwise follow each bool when the next field is pointer- or