Align README quickstart with the TRT-LLM push path

[BarneyChambers]

TLDR

The GitHub README quickstart tells people to truss push --watch a TRT-LLM Qwen config. That command exits. The rest of the guide (development URL, truss watch, "this creates production") does not match the CLI.

What

The Qwen example is an engine-builder deploy. Default truss push is published. --watch is development mode, and the CLI has rejected that for TRT-LLM since 2024.

After this change the quickstart matches the CLI and the first-model guide:

• uvx truss push (no --watch)
• Call URL is /environments/production/sync/v1
• Login sample includes the paste-vs-browser picker. --browser / --api-key skip it. BASETEN_API_KEY is for the OpenAI client, not for skipping truss login.
• Docs links use https://docs.baseten.co/... instead of repo-root paths that 404 on GitHub.
• Config includes openai-compatible and num_builder_gpus: 2, same as the first-model guide.

The live-reload / fake-promote section is gone. --watch is pointed at Customize a model.

Why

I walked the README against cli.py after #2627. The March 2026 README rewrite glued a TRT-LLM first-model config onto the custom-Python watch/promote flow. The CLI already blocked that path. The site docs were later corrected. This file was not.

How

README only. No CLI change.

Testing

No unit tests. I checked the new commands against truss/cli/cli.py (uses_trt_llm_builder + --watch exits, default publish = True) and truss/cli/remote_cli.py (auth picker). Compared the deploy/call steps to https://docs.baseten.co/examples/deploy-your-first-model.

[cretz]

Thanks for the contribution! The README did indeed grow stale here.