PyTorch cleans up stale FIXMEs: MPS indexing fix and resolved issues

PyTorch, the widely used deep learning framework (100k GitHub stars), continues its code maintenance with a small but meaningful cleanup. Contributor cyyever removed 4 stale FIXME comments from the ATen core library and related tests. These comments referenced issues #105716 and #108205, which have since been closed or resolved. Additionally, comments about code that had already been updated and a note about MPS (Metal Performance Shaders) uint8 advanced indexing no longer applying were removed, as that feature now works correctly. The pull request (#184702) was approved by core maintainer albanD.

While not a major feature addition, such housekeeping reflects the project's maturity and active maintenance. Removing outdated comments reduces confusion for new contributors and signals that known issues have been addressed. The MPS backend, which enables GPU acceleration on Apple Silicon, continues to improve. For developers using PyTorch on macOS, this means less technical debt and more reliable documentation within the codebase.