Profiling TorchNeuron (Part 1): System-level profiling on AWS Trainium.

Systematic analysis of inference performance can identify why a machine learning model is slow, particularly on specialized hardware like AWS Trainium. This guide details how to use `torch.profiler` with the TorchNeuron backend to capture and analyze detailed system-level traces. It explains how to configure the profiler, interpret the `key_averages()` output to compare CPU and device time, and use the Neuron Explorer tool for in-depth visualization. The process is demonstrated by profiling and comparing three different implementations of an RMSNorm operation to highlight performance differences.