[deanq]

What

Fixes the examples under examples/api, examples/endpoints, examples/serverless, and examples/rest_wrapper.py. examples/apps is unchanged: it matches the current SDK, and #557 already touches it.

File	Problem	Fix
endpoints/asyncio_job_request.py	Unconditional asyncio.WindowsSelectorEventLoopPolicy() raises AttributeError on macOS and Linux. await job.output() with the default timeout raises TimeoutError immediately on current releases (SDK fix in #662).	Remove the policy line. Pass an explicit timeout= to job.output().
api/create_endpoint.py	runpod/base:0.4.4 does not exist on Docker Hub.	Use runpod/mock-worker:latest, a real serverless worker.
api/create_template.py	Very old runpod/base:0.1.0.	runpod/base:2.0.0-ubuntu2404.
api/*, endpoints/*	Comments point to runpod config, which no longer exists.	Point to rp login.
endpoints/run_sync.py	except TimeoutError misses requests.exceptions.Timeout, which is raised when the HTTP wait exceeds timeout.	Catch both.
serverless/logger.py	Documented output matched neither local nor deployed output.	Document both: plain text locally, JSON when RUNPOD_ENDPOINT_ID is set.
serverless/concurrent_handler.py	Handler named async_generator_handler was not a generator; the modifier was a no-op.	A real async handler and a modifier that returns a set concurrency.
rest_wrapper.py	Used stop_pod's response for the pod id.	Keep the id from create_pod. Describe what the script does.
endpoints/health.py, run.py, run_sync.py, streaming.py	Hardcoded endpoint ids. streaming.py sent mock-worker-only input without saying so.	YOUR_ENDPOINT_ID placeholders; note which worker the streaming input targets.

Second commit: new-user pass

• examples/README.md (new): prerequisites, which folder to start with, running handlers locally, getting an endpoint id, cost and cleanup. The top-level README's Examples link now points here instead of examples/apps.
• api/*: unique names (re-running no longer collides on template names), cleanup commands, placeholder registry credentials. get_endpoints.py prints ids and names instead of every endpoint's full config, including env.
• endpoints/run.py: waits with a timeout. The bare output() call printed None while the job was queued.
• rest_wrapper.py: terminates the pod in finally, so a failed resume doesn't leave a billable pod running.

Why

None of these files run in CI, so they had drifted from the SDK. Running examples in CI is tracked separately.

Test plan

• make quality-check (2773 passed)
• serverless/logger.py: output matches the comments, both locally and with RUNPOD_ENDPOINT_ID set
• serverless/concurrent_handler.py and simple_handler.py with --test_input
• endpoints/asyncio_job_request.py, run.py, run_sync.py, health.py, streaming.py against a local fake API (RUNPOD_ENDPOINT_BASE_URL), with jobs going from IN_PROGRESS to COMPLETED
• Docker Hub returns the image tags used
• endpoints/run.py (second commit) against the fake API: prints the status, then the output
• api/* and rest_wrapper.py not run: they create billable resources. Only the signatures and the image tags were checked.