Add ANE power and profiling

[AlexCheema]

Summary

• track ANE power in system performance state and bench power usage summaries
• add ANE profiler results to event/state plumbing and dashboard aggregates
• measure ANE FP16 conv throughput plus streaming bandwidth through AppleNeuralEngine.framework

Verification

• uv run basedpyright
• uv run ruff check
• nix fmt
• uv run pytest
• npm run build
• real cluster smoke: james, mike, s13, s14 all returned ANE profiles (~6.36-6.75 TFLOPS, ~68-72 GB/s)

Note: npm run check still fails on pre-existing dashboard type errors unrelated to this branch.