Integrate Pytorch Profiler into Modulus, including StaticCapture*

[coreyjadams]

Enable Profiler in two modes in modulus:

1. Standalone, provide default configurations and both context and decorator functionality. This will let users quickly slap a profiler on their code, with useful configuration enabled.
2. Integrate into StaticCaptureTrain and StaticCaptureEval: users employing these tools should be able to capture profiles without leaving that interface.

Modulus will also provide some basic configurations for the profiling (lightweight, CPU-only, full, nvtx-annotations, etc).

Profiler must work in both distributed and non-distributed mode.