Fix invalid PromQL in logical replication PrometheusRules (#868)

Two categories of parse errors prevented the PrometheusRule from being
accepted when monitoring.enabled=true:

1. CNPGClusterLogicalReplicationErrors and CNPGClusterLogicalReplicationErrorsCritical:
   `increase()` was wrapping `max by()(A + B)[5m]`, applying a range selector
   to an aggregated instant vector, which is illegal in PromQL. Fix: apply
   `increase([5m])` to each individual metric selector before aggregating.

2. CNPGClusterLogicalReplicationStopped and CNPGClusterLogicalReplicationStoppedCritical:
   `cnpg_pg_stat_subscription_pid` (a numeric gauge) was compared with `""`
   (a string), which is not a valid PromQL binary expression. Fix: compare
   with `0` instead.

Signed-off-by: Allex1 <Allex21ro@yahoo.com>
Co-authored-by: Itay Grudev <itay@verito.digital>

Author: Allex1

charts/cluster/prometheus_rules/cluster-logical_replication_errors-critical.yaml

@@ -9,7 +9,7 @@ annotations:
     CRITICAL: High error rate indicates persistent replication issues requiring immediate attention. This could lead to significant data inconsistency or complete replication failure. Errors include both apply errors and sync errors. The subscription may stop working if errors continue.
   runbook_url: https://github.com/cloudnative-pg/charts/blob/main/charts/cluster/docs/runbooks/CNPGClusterLogicalReplicationErrors.md
 expr: |
-  label_replace(increase(max by (namespace, job, subname) (cnpg_pg_stat_subscription_apply_error_count + cnpg_pg_stat_subscription_sync_error_count)[5m]), "cluster", "$1", "job", ".+/(.+)") >= 5
+  label_replace(max by (namespace, job, subname) (increase(cnpg_pg_stat_subscription_apply_error_count[5m]) + increase(cnpg_pg_stat_subscription_sync_error_count[5m])), "cluster", "$1", "job", ".+/(.+)") >= 5
 for: 0m
 labels:
   severity: critical

charts/cluster/prometheus_rules/cluster-logical_replication_errors.yaml

@@ -9,7 +9,7 @@ annotations:
     This includes both apply errors (during normal replication) and sync errors (during initial table sync). Errors indicate data consistency issues that need immediate attention to prevent data divergence.
   runbook_url: https://github.com/cloudnative-pg/charts/blob/main/charts/cluster/docs/runbooks/{{ $alert }}.md
 expr: |
-  label_replace(increase(max by (namespace, job, subname) (cnpg_pg_stat_subscription_apply_error_count + cnpg_pg_stat_subscription_sync_error_count)[5m]), "cluster", "$1", "job", ".+/(.+)") > 0
+  label_replace(max by (namespace, job, subname) (increase(cnpg_pg_stat_subscription_apply_error_count[5m]) + increase(cnpg_pg_stat_subscription_sync_error_count[5m])), "cluster", "$1", "job", ".+/(.+)") > 0
 for: 1m
 labels:
   severity: warning

charts/cluster/prometheus_rules/cluster-logical_replication_stopped-critical.yaml

@@ -18,7 +18,7 @@ expr: |
   ) or (
     # Subscription is enabled but stuck (no worker process with significant lag)
     label_replace(max by (namespace, job, subname) (cnpg_pg_stat_subscription_enabled), "cluster", "$1", "job", ".+/(.+)") == 1
-    and label_replace(max by (namespace, job, subname) (cnpg_pg_stat_subscription_pid), "cluster", "$1", "job", ".+/(.+)") == ""
+    and label_replace(max by (namespace, job, subname) (cnpg_pg_stat_subscription_pid), "cluster", "$1", "job", ".+/(.+)") == 0
     and label_replace(max by (namespace, job, subname) (cnpg_pg_stat_subscription_buffered_lag_bytes), "cluster", "$1", "job", ".+/(.+)") / 1024^3 > 0.1
   )
 for: 15m

charts/cluster/prometheus_rules/cluster-logical_replication_stopped.yaml

@@ -17,7 +17,7 @@ expr: |
   ) or (
     # Subscription is enabled but stuck (no worker process with significant lag)
     label_replace(max by (namespace, job, subname) (cnpg_pg_stat_subscription_enabled), "cluster", "$1", "job", ".+/(.+)") == 1
-    and label_replace(max by (namespace, job, subname) (cnpg_pg_stat_subscription_pid), "cluster", "$1", "job", ".+/(.+)") == ""
+    and label_replace(max by (namespace, job, subname) (cnpg_pg_stat_subscription_pid), "cluster", "$1", "job", ".+/(.+)") == 0
     and label_replace(max by (namespace, job, subname) (cnpg_pg_stat_subscription_buffered_lag_bytes), "cluster", "$1", "job", ".+/(.+)") / 1024^3 > 0.1
   )
 for: 5m