fix(pgtype): return error instead of panicking on malformed interval text

scanPlanTextAnyToIntervalScanner read timeParts[0][0] without checking that
the hours segment is non-empty. An interval text value whose time component
starts with a colon, such as ":05:06", splits into an empty hours segment and
panics with "index out of range [0] with length 0".

Guard the byte access so an empty hours segment falls through to
strconv.ParseInt, which already returns a "bad interval hour format" error.
This matches the malformed-text handling added for the geometric types.

Author: greymoth-jp

pgtype/interval.go

@@ -232,7 +232,7 @@ func (scanPlanTextAnyToIntervalScanner) Scan(src []byte, dst any) error {
 		}
 
 		var negative bool
-		if timeParts[0][0] == '-' {
+		if len(timeParts[0]) > 0 && timeParts[0][0] == '-' {
 			negative = true
 			timeParts[0] = timeParts[0][1:]
 		}

pgtype/interval_malformed_test.go

@@ -0,0 +1,19 @@
+package pgtype_test
+
+import (
+	"testing"
+
+	"github.com/jackc/pgx/v5/pgtype"
+)
+
+// A malformed interval whose time component has an empty hours segment
+// (for example ":05:06") must return an error instead of panicking with an
+// index out of range on timeParts[0][0].
+func TestIntervalScanMalformedTimeReturnsError(t *testing.T) {
+	for _, src := range []string{":05:06", "1 day :30:00"} {
+		var v pgtype.Interval
+		if err := v.Scan(src); err == nil {
+			t.Errorf("Scan(%q): expected error, got nil", src)
+		}
+	}
+}