fix: stop actual-budget-exporter probes from hitting login-triggering /metrics

readinessProbe/livenessProbe hit /metrics on top of Prometheus's own
scrapes; exporter re-logs into actual_server on every call (no session
reuse), tripped actual_server's login rate limit (too-many-requests).
Switch probes to tcpSocket and stretch scrape interval to 5m.

apps/actual-budget-exporter/chart/templates/deployment.yaml

@@ -41,17 +41,19 @@ spec:
                   key: ACTUAL_E2E_PASSWORD_1
           resources:
             {{- toYaml .Values.resources | nindent 12 }}
-          # No documented dedicated health endpoint; /metrics is the route
-          # that's always up once the exporter has started.
+          # /metrics does a full login against actual_server on every hit
+          # (see actual-api-service.js: initializeApi() runs per getMetrics()
+          # call) — an httpGet probe against it every 10-20s was stacking on
+          # top of Prometheus's own scrapes and tripped actual_server's
+          # login-attempt rate limit ("too-many-requests"). Probe TCP only;
+          # let Prometheus be the sole thing that ever calls /metrics.
           readinessProbe:
-            httpGet:
-              path: /metrics
+            tcpSocket:
               port: {{ .Values.service.port }}
             initialDelaySeconds: 5
             periodSeconds: 10
           livenessProbe:
-            httpGet:
-              path: /metrics
+            tcpSocket:
               port: {{ .Values.service.port }}
             initialDelaySeconds: 15
             periodSeconds: 20

apps/actual-budget-exporter/chart/values.yaml

@@ -34,4 +34,6 @@ resources:
 
 serviceMonitor:
   enabled: true
-  interval: 30s
+  # Each scrape does a full login against actual_server (no session reuse in
+  # this exporter) — kept long to avoid tripping its login rate limit.
+  interval: 5m