[Docs] Triton installation

[iwr-redmond]

This issue can be closed when #5 is completed.

Installing Triton is possible on Windows with CUDA and Intel XPUs and on Linux with CUDA, AMD ROCm, and Intel XPUs. There does not appear to be a Triton distribution for Apple Silicon (MPS).

• On Windows with CUDA, use the command: pip install triton-windows
• On Linux with CUDA, Triton is installed automatically with Torch
• On Linux with AMD ROCm, use the command: torchruntime install pytorch-triton-rocm
• On Windows and Linux with an Intel XPU, use the command: torchruntime install pytorch-triton-xpu

Windows with CUDA support is provided by a third-party package.

[iwr-redmond]

There is an beta-quality build of Triton 3.3 for Apple Silicon (MPS) here but it is unmaintained and I reckon not appropriate for general use.