Add GPU-side Gumbel-max sampling for CUDA graph compatibility

[Gasoonjia]

This PR replaces cpu sampler with CUDA sampler and fuse sampler with forward method to both eliminate unnecessary data transfer and improve sampling efficient. Decode performance increases from 113.8 token/s to 119.5 token/s

Once we land the device support pipeline, we should decompose the forward method with sampling.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18844

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• Rolling out OSDC (ARC) runners on pull workflow for PyTorch trunk commits

❌ 7 New Failures, 4 Unrelated Failures

As of commit e207ffc with merge base 2eaa16c ():

NEW FAILURES - The following jobs have failed:

• pull / test-multimodal-linux (gemma3-4b) / linux-job (gh)
  RuntimeError: Command docker exec -t 22b0340d850e4abaafab66573e4d9371c47326175fe6074840bcf98d5159a779 /exec failed with exit code 134
• Test CUDA Builds / test-model-cuda-e2e (nvidia, parakeet-tdt, non-quantized) / linux-job (gh)
  RuntimeError: Command docker exec -t c652874dbebf67770c74ca4d55381290591b7d47cdee418781497d46dd317647 /exec failed with exit code 1
• Test CUDA Builds / test-model-cuda-e2e (nvidia, parakeet-tdt, quantized-int4-tile-packed) / linux-job (gh)
  RuntimeError: Command docker exec -t c0ea9f1992e40729fa2ea12141b35ebcc00e34aa6ab32322017f976526d8bfb5 /exec failed with exit code 1
• Test CUDA Builds / unittest-cuda / linux-job (gh)
  examples/models/qwen3_5_moe/test_turboquant.py::TestTurboQuant::test_kv_cache_state_matters
• Test CUDA Windows Export and E2E / export-model-cuda-windows-artifact (facebook, dinov2-small-imagenet1k-1-layer, non-quantized) / linux-job (gh)
  RuntimeError: Command docker exec -t d330e9ca70092c35df30284c17bfe01dd585cc526719cfb7bcfbd9ec09c2b3e6 /exec failed with exit code 1
• Test CUDA Windows Export and E2E / export-model-cuda-windows-artifact (nvidia, parakeet-tdt, non-quantized) / linux-job (gh)
  RuntimeError: Command docker exec -t 41b875b457330a870c2062640f6c5d0dca9436b60061ecd741ba610abb84f853 /exec failed with exit code 1
• Test CUDA Windows Export and E2E / export-model-cuda-windows-artifact (nvidia, parakeet-tdt, quantized-int4-weight-only) / linux-job (gh)
  RuntimeError: Command docker exec -t b5092eef6d8a446f9fe212cc307bb25519913ed64331ccbabc64b017eb3805e9 /exec failed with exit code 1

FLAKY - The following jobs failed but were likely due to flakiness present on trunk:

• Test ARM Backend / test-arm / test-backend-linux (arm_tosa_fp, models) / linux-job (gh) (detected as infra flaky with no log or failing log classifier)
• Test ARM Backend / test-arm / test-backend-linux (arm_vgf_fp, models) / linux-job (gh) (detected as infra flaky with no log or failing log classifier)

BROKEN TRUNK - The following jobs failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• pull / unittest / windows / windows-job (gh) (trunk failure)
  ##[error]The operation was canceled.
• pull / unittest-editable / windows / windows-job (gh) (trunk failure)
  ##[error]The operation was canceled.

This comment was automatically generated by Dr. CI and updates every 15 minutes.