fix: prevent infinite loading screen by adding sweep timeouts

The dashboard would hang indefinitely on the loading screen because:

1. `bls.mjs` used a raw `fetch()` without any timeout/AbortSignal —
   if the BLS API was slow or unresponsive, it would block forever.

2. `runSource()` in `briefing.mjs` had no per-source timeout, so a
   single hanging API could stall the entire sweep indefinitely.

3. `server.mjs` loaded cached `latest.json` via a fire-and-forget
   promise (`.then()`) instead of `await`, meaning the dashboard
   never received the cached data before the sweep started.

4. `loading.html` relied solely on SSE for redirect — if the SSE
   connection missed the update event, the page would never redirect.

Changes:
- Add 15s AbortController timeout to BLS `getSeries()` fetch call
- Add 30s per-source timeout via `Promise.race()` in `runSource()`
- Await `synthesize()` when loading cached data so the dashboard
  serves instantly on restart when `runs/latest.json` exists
- Add 5s fallback polling to loading page alongside SSE

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

apis/sources/bls.mjs

@@ -37,11 +37,15 @@ export async function getSeries(seriesIds, opts = {}) {
   if (apiKey) payload.registrationkey = apiKey;
 
   try {
+    const controller = new AbortController();
+    const timer = setTimeout(() => controller.abort(), 15000);
     const res = await fetch(base, {
       method: 'POST',
       headers: { 'Content-Type': 'application/json' },
       body: JSON.stringify(payload),
+      signal: controller.signal,
     });
+    clearTimeout(timer);
     return await res.json();
   } catch (e) {
     return { error: e.message };