Add Support for Kimi K2.5 Vision

[quic-mamta]

Summary

This PR adds QEff support for Kimi K2.5 Vision using a ConditionalGeneration flow designed for Dual-QPC execution, and updates our Transformers dependency to align with the required upstream APIs.

What’s included

• Added QEff Kimi K2.5 multimodal model support:
• New Kimi K2.5 config classes (text + vision).
• New Kimi K2.5 model implementation with:
• vision tower + projector integration,
• language decoder integration via DeepSeek-V3,
• multimodal input merging and generation path,
• export/runtime helpers (dummy inputs, dynamic axes, specializations, output naming).
• Added a Kimi K2.5 vision export example for Dual-QPC
• Added parity checks for vision and language models.

Validation

• Verified model wiring and transform registration are in place for Kimi K2.5 conditional generation flow.
• Verified export example script supports reduced-layer / expert-subset setup for iterative validation.

Example Script -
python examples/kimi_k2/export_kimi_k25_vision.py

Tests -
python examples/kimi_k2/test_kimi_k25.py

[github-advanced-security]

You are seeing this message because GitHub Code Scanning has recently been set up for this repository, or this pull request contains the workflow file for the Code Scanning tool.

What Enabling Code Scanning Means:

• The 'Security' tab will display more code scanning analysis results (e.g., for the default branch).
• Depending on your configuration and choice of analysis tool, future pull requests will be annotated with code scanning analysis results.
• You will be able to see the analysis results for the pull request's branch on this overview once the scans have completed and the checks have passed.

For more information about GitHub Code Scanning, check out the documentation.