feat: Schritt 4 — Locked scheduler (global GPU lock, queue, stats)

Replaces the MVP Passthrough with scheduler.Locked: a capacity-1 channel
serialises every consumer's GPU work end-to-end. main.go switches to it.

Behavioural contract:
- Jobs that arrive while another job holds the GPU block on the channel
  until the holder finishes. Context cancellation aborts the wait
  cleanly (no leaked tokens, queue depth decremented).
- Stats track queue_depth, in_flight, total_jobs, last_wait_ms,
  last_run_ms, oldest_queued — surfaced through /v1/status.
- One lock for ALL consumers (not per-consumer): the design (§4.3) is
  explicit that grobgranular > GPU-stream-granular on single-GPU
  single-user hardware. mvoice + ollama + comfyui never run truly
  concurrently any more, which is the whole point — that's what
  produced the CUDA-OOM under load.

Tests:
- 5 goroutines hammer the scheduler concurrently → max in-flight = 1.
- Cancellation while parked on the lock returns ctx.Err() and frees
  the queue slot.
- Stats reflect in-flight + queue-depth transitions correctly.
- Race detector clean.

Schritt 5 will compose this with VRAM-pressure eviction: before
acquiring the lock, check if the target consumer's resident cost fits
under the current GPU headroom; if not, unload the LRU non-coexistent
consumer first.

Refs: m/mGPUmanager#1 (Schritt 4).

cmd/mgpumanager/main.go

@@ -61,7 +61,9 @@ func main() {
 
 	reg := registry.New(cfg, logger.With("component", "registry"))
 	gpuPoller := gpu.NewPoller(cfg.GPU.PollInterval(), logger.With("component", "gpu"))
-	sched := scheduler.NewPassthrough(reg)
+	// Phase 1 always runs a single-slot global GPU lock. Schritt 5's
+	// eviction-aware scheduler wraps this same lock with VRAM pressure logic.
+	sched := scheduler.NewLocked(reg, 1)
 
 	go reg.Run(ctx)
 	go gpuPoller.Run(ctx)