feat(gomsg): pymsg/libmsg 协议的 Go 实现，waf logclient 抽取接入

- 新增 gomsg/：帧协议（4B 大端长度 + UTF-8，64MiB 上限）、MsgClient/MsgServer、
  RpcClient/RpcServer（_route 分发、错误脱敏、panic 恢复、json.Number 保精度）、
  AsyncQueue[T]、LogClient（集中式日志，级别过滤 + fail-safe），线协议与 Python/C++ 字节级兼容
- waf/internal/logclient 改为 gomsg 薄适配层（336→20 行），main/engine/proxy 零改动
- gomsg 26 项单测全绿（-race）；waf 编译通过

gomsg/asyncqueue_test.go

@@ -0,0 +1,120 @@
+package gomsg
+
+import (
+	"sync"
+	"testing"
+	"time"
+)
+
+func collectBatches(batches *[][]int, mu *sync.Mutex) func([]int) {
+	return func(b []int) {
+		mu.Lock()
+		*batches = append(*batches, append([]int(nil), b...))
+		mu.Unlock()
+	}
+}
+
+func TestAsyncQueueFlushOnBatchSize(t *testing.T) {
+	var batches [][]int
+	var mu sync.Mutex
+	q := NewAsyncQueue[int](collectBatches(&batches, &mu), AsyncQueueConfig{
+		BatchSize:     5,
+		FlushInterval: time.Hour, // 只测批量阈值，不测定时
+	})
+	for i := 1; i <= 5; i++ {
+		if !q.Push(i) {
+			t.Fatal("push should succeed")
+		}
+	}
+	deadline := time.Now().Add(2 * time.Second)
+	for {
+		mu.Lock()
+		n := len(batches)
+		mu.Unlock()
+		if n >= 1 {
+			break
+		}
+		if time.Now().After(deadline) {
+			t.Fatal("batch not flushed in time")
+		}
+		time.Sleep(10 * time.Millisecond)
+	}
+	mu.Lock()
+	defer mu.Unlock()
+	if len(batches[0]) != 5 || batches[0][4] != 5 {
+		t.Fatalf("batch = %v", batches[0])
+	}
+	q.Shutdown()
+}
+
+func TestAsyncQueueFlushOnInterval(t *testing.T) {
+	var batches [][]int
+	var mu sync.Mutex
+	q := NewAsyncQueue[int](collectBatches(&batches, &mu), AsyncQueueConfig{
+		BatchSize:     100,
+		FlushInterval: 100 * time.Millisecond,
+	})
+	if !q.Push(42) {
+		t.Fatal("push should succeed")
+	}
+	time.Sleep(300 * time.Millisecond)
+	mu.Lock()
+	n := len(batches)
+	mu.Unlock()
+	if n < 1 {
+		t.Fatal("expected interval flush")
+	}
+	q.Shutdown()
+}
+
+func TestAsyncQueueShutdownDrains(t *testing.T) {
+	var batches [][]int
+	var mu sync.Mutex
+	q := NewAsyncQueue[int](collectBatches(&batches, &mu), AsyncQueueConfig{
+		BatchSize:     100,
+		FlushInterval: time.Hour,
+	})
+	for i := 1; i <= 3; i++ {
+		q.Push(i)
+	}
+	q.Shutdown()
+	mu.Lock()
+	defer mu.Unlock()
+	if len(batches) != 1 || len(batches[0]) != 3 {
+		t.Fatalf("expected drained batch [1 2 3], got %v", batches)
+	}
+}
+
+func TestAsyncQueueDropOnFull(t *testing.T) {
+	var batches [][]int
+	var mu sync.Mutex
+	q := NewAsyncQueue[int](collectBatches(&batches, &mu), AsyncQueueConfig{
+		BatchSize:     1000,
+		FlushInterval: time.Hour,
+		MaxQueueSize:  4,
+	})
+	// 后台协程可能已取走部分，填满队列直到 Push 返回 false
+	dropped := false
+	for i := 0; i < 100; i++ {
+		if !q.Push(i) {
+			dropped = true
+			break
+		}
+	}
+	if !dropped {
+		t.Fatal("expected drop when queue full")
+	}
+	d, _ := q.Stats()
+	if d == 0 {
+		t.Fatal("dropped counter should be > 0")
+	}
+	q.Shutdown()
+}
+
+func TestAsyncQueueRejectAfterShutdown(t *testing.T) {
+	q := NewAsyncQueue[int](func([]int) {}, AsyncQueueConfig{})
+	q.Shutdown()
+	if q.Push(1) {
+		t.Fatal("push after shutdown should fail")
+	}
+}