fix(metrics): include additional check metric labels in canary_check_invalid_count

When --metric-labels-allowlist is configured, OpsInvalidCount was registered
with only the base 8 labels while Record() called it with checkMetricLabels
that include the additional labels. This caused Prometheus label mismatch
panics which were silently swallowed by recover(), resulting in invalid
check metrics not being recorded.

Fix: use checkLabels (which appends AdditionalCheckMetricLabels) for
OpsInvalidCount, matching all other counters that already use it.

Author: adityathebe

pkg/metrics/metrics.go

@@ -62,7 +62,7 @@ func SetupMetrics() {
 			Name: "canary_check_invalid_count",
 			Help: "The total number of invalid checks",
 		},
-		[]string{"type", "endpoint", "canary_name", "canary_namespace", "owner", "severity", "key", "name"},
+		checkLabels,
 	)
 
 	CanaryCheckInfo = prometheus.NewGaugeVec(