Fix CPU offload regressions and optimize UV unwrapping for dense meshes

[undergroundrap]

Title: Fix CPU offload regressions and optimize UV unwrapping

Description:
This PR addresses several issues detected when running in low-VRAM/CPU-offload modes:

1. Bug Fix (AttributeError): Added self.components to Hunyuan3DDiTPipeline to support accelerate offloading hooks.
2. Bug Fix (Device Mismatch): Ensured sigma tensors in schedulers are moved to the correct device during sampling.
3. Optimization: Added optional mesh decimation to mesh_uv_wrap to prevent performance hangs on dense meshes (1M+ faces).
4. Dependency: Added fast-simplification to requirements.txt for mesh optimization.

[tencent-adm]

Thank you for your submission, we really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[undergroundrap]

I've submitted my signed CLA via email to TencentCLA@tencent.com, as I encountered issues with the online signing portal. Please let me know if you need any further information to verify my status. Thank you!

[undergroundrap]

I've set a default limit of 80,000 faces to ensure the UV warp completes on consumer hardware (RTX 3070/4070). If we want to support 1M+ faces for enterprise users, we could move this 'max_faces' to a config variable.