[pd-hung]

Summary

Adds cuTile support to the profiler client. Stacked on #499.

• tilelens.trace("profiler", frontend="cutile") and the tile-profiler CLI work for cuTile kernels. Under the CLI, ct.launch runs through the CPU interpreter, and each kernel is traced once so statistics accumulate across launches, like the Triton CLI.
• Reports the same statistics as for Triton: load/store bytes, masked elements, and for-loop lengths. For cuTile, "masked" counts the padded out-of-bounds elements of partial tiles, since tiled loads have no user mask.
• For-loops are found by rewriting kernel and helper source, like Triton and NKI (tilelens/core/frontend/cutile_transform.py).
• Load/store skipping returns cuTile-shaped zeros for loads and matmuls and drops stores. Atomics are never skipped, matching Triton.
• Block sampling works as in Triton. The AMD buffer-load check is turned off for cuTile.
• CUDA tensors run on host copies that are written back after the launch, like Triton's interpreter.

Test Plan

New tests/cutile/test_cutile_profiler.py: byte and mask counts, loops, skipping, atomics, block sampling, and the CLI.
uv run pytest tests/

Related Issues

Breaking Changes

Checklist

• I added tests to all new functionality I added/bugs I fixed.
• I verified that a human has reviewed all code in this PR.
• I ran npm run build:frontend if the PR modified any TypeScript code.
• I made sure that my code is well documented (comments explaining strange code, docstrings for functions, website modified if new functionality added).

[github-actions]

Performance Benchmark

Benchmark	main (min)	PR (min)	Change	Samples
gemm	N/A	N/A	N/A	0 / 0
gemm_oob	N/A	N/A	N/A	0 / 0
indirect_load	N/A	N/A	N/A	0 / 0
nested_loop	N/A	N/A	N/A	0 / 0
block_pointer_loop_advance	N/A	N/A	N/A	0 / 0
liger_jsd	N/A	N/A	N/A	0 / 0
flaggems_layernorm	N/A	N/A	N/A	0 / 0
swiglu	0.163s	0.163s	-0.0%	20 / 20
cross_entropy	N/A	N/A	N/A	0 / 0
fused_linear_jsd	N/A	N/A	N/A	0 / 0
Total	N/A	N/A	N/A	N/A

Iterations: 1 warmup + 20 measured
Samples are shown as main / PR; long pytest benchmarks may use fewer samples.