MLA attention fallback

[jloftin-nv]

Description

When the SDPA validator declines (e.g. MLA K/V head-dim mismatch), the node is left as an unconvertible PyTorch op even though decompose_attention=True would work. Users must know to set that flag by hand.

After the causal-attention pass, auto-decompose SDPA nodes the validator declines so MLA models do not take one graph break per layer.

Type of change

Please delete options that are not relevant and/or add your own.

• New feature (non-breaking change which adds functionality)

Checklist:

• [ X] My code follows the style guidelines of this project (You can use the linters)
• [ X] I have performed a self-review of my own code
• [ X] I have commented my code, particularly in hard-to-understand areas and hacks
• [ X] I have made corresponding changes to the documentation
• [ X] I have added tests to verify my fix or my feature
• [ X] New and existing unit tests pass locally with my changes
• [ X] I have added the relevant labels to my PR in so that relevant reviewers are notified