Document context_model using SumContextEncoder

Retrain context_model with SumContextEncoder-based network (time+label folded into one context vector)

Update README.md

Document context_model as coordinates-only with label context

Retrain context_model: coordinates-only modality, label as context (drop index head)

Update README.md

Update README.md

Document the conditioning-tutorial checkpoints

Add conditioning-tutorial checkpoint: context_model

Add conditioning-tutorial checkpoint: joint_model

initial commit