feat(03-01): extend TrafficClass to 14 classes with hash-bucketed unknowns

- Replace ClassUnknown with ClassUnknown1-4 (unknown-1 through unknown-4)
- AllClasses() now returns 14 elements
- Add hashBucket() function routing unrecognized traffic deterministically
- Update synth/config.go: NumLayers=14, 4 dissonant unknown tone configs at 862-1047 Hz
- Add TestAllClassesCount and TestHashBucketDistribution tests
- Add TestClassFreqConfigsComplete and TestNumLayersMatchesAllClasses to config_test.go
- Update TestNewBankHas14Layers, TestMixerNoClip for 14 classes
- All classify and synth tests pass

synth/bank_test.go

@@ -7,10 +7,10 @@ import (
 	"github.com/netsynth/netsynth/classify"
 )
 
-func TestNewBankHas11Layers(t *testing.T) {
+func TestNewBankHas14Layers(t *testing.T) {
 	b := NewBank(1.0)
-	if len(b.layers) != 11 {
-		t.Errorf("NewBank() has %d layers, want 11", len(b.layers))
+	if len(b.layers) != 14 {
+		t.Errorf("NewBank() has %d layers, want 14", len(b.layers))
 	}
 	// Verify each class has exactly one layer
 	for _, class := range classify.AllClasses() {
@@ -76,13 +76,13 @@ func TestRenderWindowNonZeroWithTraffic(t *testing.T) {
 func TestMixerNoClip(t *testing.T) {
 	b := NewBank(0.01) // fast EMA to quickly ramp up to near-max amplitude
 	counts := make(map[classify.TrafficClass]int64)
-	// All 11 classes at max count — worst-case mixing scenario
+	// All 14 classes at max count — worst-case mixing scenario
 	for _, class := range classify.AllClasses() {
 		counts[class] = 1000
 	}
 	snap := classify.WindowSnapshot{
 		Counts:       counts,
-		TotalPackets: 11000,
+		TotalPackets: 14000,
 		WindowIndex:  0,
 	}
 	// Render multiple windows to let EMA converge