feat: whisper-side cancellation + SSE-triggered retry

- Add cancelJob() to whisper.ts: sends DELETE /jobs/:id to the whisper
  server (best-effort, errors silently ignored)
- DELETE /api/jobs/[id] now calls cancelJob() when cancelling an active
  job that has a whisperJobId, stopping GPU use immediately
- Webhook handler guards against locally-cancelled jobs: returns ok early
  so whisper's late completion cannot overwrite cancelled status or send
  a phantom 'Transcript ready' notification
- Replace blind sleep(Retry-After + 1s) in submitJob() with
  waitForModelReady(): subscribes to /model/events SSE and proceeds as
  soon as state:ready arrives; falls back to the Retry-After timeout if
  SSE is unreachable or closes without model_ready
- Refactor retry tests to use URL-aware makeJobFetch() helper; add 7 new
  tests (3 SSE-triggered retry, 3 cancelJob, 1 webhook cancelled-guard)
  — 144/144 passing

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

src/routes/api/jobs/[id]/+server.ts

@@ -1,5 +1,6 @@
 import { json, error } from '@sveltejs/kit';
 import { getJob, setJobStatus, deleteJob } from '$lib/server/db.js';
+import { cancelJob } from '$lib/server/whisper.js';
 import { rm } from 'fs/promises';
 
 export async function GET({ params }) {
@@ -17,6 +18,8 @@ export async function DELETE({ params }) {
 	if (ACTIVE.has(job.status)) {
 		// Cancel active job (keeps DB record)
 		setJobStatus(params.id, 'cancelled', 0);
+		// Best-effort: tell whisper to drop the queued job so it stops using GPU
+		if (job.whisperJobId) cancelJob(job.whisperJobId).catch(() => {});
 	} else {
 		// Hard-delete terminal job + clean up output files
 		deleteJob(params.id);

src/routes/api/webhook/[jobId]/+server.ts

@@ -12,6 +12,11 @@ const jobId = params.jobId;
 const job = getJob(jobId);
 if (!job) throw error(404, 'Job not found');
 
+// Discard the result if the job was cancelled locally while whisper was running
+if (job.status === 'cancelled') {
+return json({ ok: true });
+}
+
 const whisperJob = (await request.json()) as WhisperJob;
 
 if (whisperJob.status === 'failed' || whisperJob.status === 'cancelled') {