datadog: sanitize characters in Latin-1 supplement

Add test cases to reproduce the panic, and then fix the panic.

Author: kevinburkesegment

datadog/serializer.go

@@ -87,7 +87,7 @@ func (s *serializer) AppendMeasures(b []byte, _ time.Time, measures ...stats.Mea
 var accentMap [256]byte
 
 // valid[byte] = 1 if the ASCII char is allowed, 0 otherwise.
-var valid = [128]bool{
+var valid = [256]bool{
 	'.': true, '-': true, '_': true,
 }
 

datadog/serializer_test.go

@@ -298,6 +298,12 @@ func TestAppendSanitizedMetricName(t *testing.T) {
 		{"", "naïve🌍test", "naive_test"},             // ï -> i, emoji -> _
 		{"", "umlauts.🤡München", "umlauts._Munchen"}, // emoji + German umlauts
 
+		// Unmapped Latin-1 Supplement characters (regression test for panic)
+		{"", "test÷metric", "test_metric"},                 // ÷ (division sign, U+00F7, byte 247)
+		{"", "value×count", "value_count"},                 // × (multiplication sign, U+00D7, byte 215)
+		{"", "price¤amount", "price_amount"},               // ¤ (currency sign, U+00A4)
+		{"prefix_", "data÷by×time", "prefix_data_by_time"}, // multiple unmapped chars with prefix
+
 		// empty or only illegal
 		{"", "", "_unnamed_"},
 		{"", "!!!", "_truncated_"},