feat: add structured backend logging

Collector failures, service-API errors, WS connection lifecycle, and
config reload/discovery events were previously invisible outside the
browser (or, for a few config/discovery cases, logged with an
inconsistent ad-hoc console.error). Add a small scoped logger
(lib/logger.ts, level via LOG_LEVEL) and wire it through server.ts,
the scheduler, service collector, config loader/effective store, and
the WS server so operators can see failures via `docker logs`.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

lib/collectors/service.ts

@@ -8,6 +8,9 @@ import { collectN8n } from "./services/n8n";
 import { collectUmami } from "./services/umami";
 import { collectHeadscale } from "./services/headscale";
 import { collectTraefikStat } from "./services/traefik";
+import { createLogger } from "@/lib/logger";
+
+const log = createLogger("service");
 
 function collectServiceStat(widget: ServiceWidget): Promise<{ stats: ServiceStat[] }> {
   switch (widget.service) {
@@ -34,10 +37,11 @@ export async function collectService(widget: ServiceWidget): Promise<ServiceWidg
   // A failed service-API call (bad token, service down) must not take out the
   // docker health readout too - the two are independent failure modes and
   // merging them into one card should not let one mask the other.
-  const statPromise = collectServiceStat(widget).catch((err) => ({
-    stats: [] as ServiceStat[],
-    statError: (err as Error).message,
-  }));
+  const statPromise = collectServiceStat(widget).catch((err) => {
+    const message = (err as Error).message;
+    log.warn("service API call failed", { service: widget.service, containerName: widget.containerName, error: message });
+    return { stats: [] as ServiceStat[], statError: message };
+  });
 
   const [docker, statResult] = await Promise.all([dockerPromise, statPromise]);
   return { docker, ...statResult };