adding TapNet model

[JustinStephan]

Summary

• Add Time Series Attentional Prototype Network (pyhealth.models.tapnet.TapNet, TapNetLayer) that combines temporal conv, attention pooling, and learnable prototypes per feature stream, with optional embedding and prototype-attention outputs.
• Register TapNet in pyhealth.models.__init__.
• Add unit tests (tests/core/test_tapnet.py) covering init metadata, forward shapes/outputs, backward gradients, optional embeddings/prototype attention, and custom hyperparameters.
• Add top-level doc TAPNET_README.rst with usage snippet and test instructions.

Testing

• PYTHONPATH=. python -m pytest tests/core/test_tapnet.py
  • Ensure pytest is installed in the active env.
  • If ~/.cache/pyhealth/ is not writable, set a writable HOME before running.