Add ActivityType::OVERHEAD to kXpuTypes so that profiling with ProfilerActivity.XPU enables PTI's collection-overhead view, matching the CUDA backend (kCudaTypes already lists OVERHEAD). The libkineto xpupti plugin already handles the records; this was the only missing wiring on the PyTorch side.

OVERHEAD events render on a dedicated track (deviceId == -1) and attribute the profiler's own collection cost, rather than reducing it.

Add a Python integration test asserting the overhead category shows up in the exported chrome trace with well-formed metadata.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/187835
Approved by: https://github.com/ryanzhang22

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>