Add inference quickstart guide

[favyen2]

This covers how to initialize the model, get a satellite image, and compute embeddings on it.

This is intended to be lightweight but we can have a more detailed one in rslearn that uses rslearn to get an image time series to apply OlmoEarth on.

---

Note

Adds a concise quickstart guide showing how to initialize OlmoEarth, fetch a Sentinel-2 image, and compute embeddings with example code.

• Docs:
  • Add docs/inference_quickstart.md quickstart guide for running inference with OlmoEarth.
    • Environment setup (Python 3.12 with uv) and model initialization via load_model(ModelID.OLMOEARTH_V1_BASE).
    • Sentinel-2 L2A data acquisition via Copernicus Browser or direct wget link; unzip instructions.
    • Preprocessing with rasterio/WarpedVRT to read bands at 10m, stack, and reshape; normalization with Normalizer(Strategy.COMPUTED).
    • Inference example creating MaskedOlmoEarthSample, running model.encoder(...), and pooling features to obtain embeddings.

^{Written by Cursor Bugbot for commit 237b8b5. This will update automatically on new commits. Configure here.}

[Hgherzog]

LGTM! Thanks for adding this