Document platform and CPU/GPU install notes

[Bortlesboat]

Summary

• add platform notes for supported Python versions and OS coverage
• clarify CPU-only vs GPU install expectations
• point users to the official PyTorch selector for CPU/CUDA wheel selection

Fixes #158.

Verification

• git diff --check

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[gemini-code-assist]

Code Review

This pull request adds a 'Platform and hardware notes' section to the README, detailing Python version support, OS compatibility, and installation instructions for both CPU and GPU environments. The review feedback suggests refining the hardware terminology to be more inclusive of macOS (MPS) rather than focusing solely on NVIDIA (CUDA) and recommends specifying a sample limit for CPU-based inference to provide better guidance for users.

I am having trouble creating individual review comments. Click here to see my feedback.

README.md (45)

The mention of 'CUDA' as a requirement for GPU usage is specific to Linux and Windows with NVIDIA hardware. Since macOS is a supported platform (using MPS), it would be more accurate to use more inclusive terminology or mention MPS explicitly.

- **Operating systems**: Linux, macOS, and Windows are covered by the test matrix. GPU usage depends on your local PyTorch setup and hardware drivers (e.g., CUDA or MPS).

README.md (46)

For consistency with the 'Quick Start' section and the library's internal settings, it is recommended to specify the sample limit for CPU inference (typically ≲1000 samples).

- **CPU-only install**: `pip install tabpfn` works without a GPU, but CPU inference is intended for small datasets (≲1000 samples). If you need a CPU-only PyTorch wheel, install PyTorch from the [official selector](https://pytorch.org/get-started/locally/) before installing TabPFN.

README.md (47)

Similar to the OS notes, the 'GPU install' instructions are currently NVIDIA-centric. It might be helpful to generalize the requirement for hardware-specific PyTorch builds to include macOS users.

- **GPU install**: Install a PyTorch build that matches your hardware (e.g., specific CUDA version) from the [official selector](https://pytorch.org/get-started/locally/), then install TabPFN with `pip install tabpfn`.