feat: scaffold load balancing worker as a template repo

[TimPietruskyRunPod]

Makes worker-load-balancing a proper template repo, mirroring the structure of worker-template (the queue-balanced one) while staying a load balancing worker.

Context: requested by Owen — turn this into a template repo similar to worker-template.

What changed

File	What
.runpod/hub.json (new)	Runpod Hub listing config. Sets "endpointType": "LB" so the Hub deploys this as a load balancer, not a queue worker. Exposes PORT/PORT_HEALTH as configurable env vars. GPU defaults (ADA_24) to match worker-template.
.runpod/tests.json (new)	Sample input for Hub validation.
README.md	Rewritten in template style: banner → description → Hub badge → "Use this template" → Customizing → Testing Locally → Deploying → Further Info. Keeps the LB-specific port config + curl examples.
public/worker-load-balancing_banner.png (new)	Placeholder banner — swap for an official design asset when available.
LICENSE (new)	MIT, matching worker-template.
app.py	Added template-guidance comments (where to load your model, /ping status codes, adding routes). Logic unchanged.
.gitignore	Expanded to the standard Python set.
requirements.txt	Bumped fastapi/uvicorn/pydantic off the 2023 pins (pydantic==1.10.7) that crash on modern Python.

Verified

Booted the server locally on the updated deps — /ping → 200, /generate, /stats, and / all respond correctly. Both JSON configs parse.

Follow-ups (can't be done in code)

• Flip the Template repository toggle in repo Settings → General (enables the "Use this template" button).
• Hub badge/listing goes live once the repo is published to the Runpod Hub.
• Confirm the Hub tests.json runner exercises LB endpoints correctly (LB workers don't take queue job input) — the file is schema-valid and harmless either way.