training: move to dedicated subprocess with ZMQ communication

- Add training_worker.py: long-lived subprocess that handles GPU training
  work, owns HF model wrapper (views into vLLM GPU memory), Apollo
  optimizer, and checkpoint sync

- train_router.py: now forwards /train requests via async ZMQ instead of
  running training in-process. Adds /checkpoint and /train/status endpoints

- export_hook.py: store model_path in __metadata__ so training worker can
  find it without cross-process communication

- This fixes two bugs:
  1. Process boundary issue - model_path was set in worker process but
     needed in API server process
  2. Blocking event loop - training blocked vLLM's async event loop

Architecture: vLLM API server <-> ZMQ <-> training subprocess
The subprocess loads IPC handles once, creates views into vLLM's GPU
memory, and handles training requests without blocking inference.

Co-Authored-By: Proof of Concept <poc@bcachefs.org>

training/pyproject.toml

@@ -11,6 +11,7 @@ dependencies = [
     "torch",
     "aiohttp",
     "safetensors",
+    "pyzmq",
 ]
 
 [project.optional-dependencies]
@@ -21,6 +22,7 @@ apollo = "apollo_plugin:register"
 
 [project.scripts]
 apollo-checkpoint = "apollo_plugin.checkpoint_sync:main"
+apollo-worker = "apollo_plugin.training_worker:main"
 
 [tool.setuptools.packages.find]
 where = ["."]