Cache RoPE freqs on device to avoid repeated CPU-GPU copy in QwenImage (#13406)
* Cache RoPE freqs on device to avoid repeated CPU-GPU copy in QwenImage
* Apply style fixes
* use lru_cache_unless_export
---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: dg845 <58458699+dg845@users.noreply.github.com>
Co-authored-by: Sayak Paul <spsayakpaul@gmail.com>