split index.Tensor converter for bool vs int indexing

[wenbingl]

Description

Add index_has_bool_indices validator and register two separate converters for torch.ops.aten.index.Tensor:

• Integer indexing (HIGH priority, no output allocator): output shape is deterministic based on index tensor shape.
• Boolean indexing (requires output allocator): uses nonzero() internally, producing data-dependent output shapes.

Fixes # (issue)
support cuda-graph feature for some beam search based model which failed due to output-allocator limit

Type of change

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

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Checklist:

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

[wenbingl]

@narendasan , please take a look whether this PR is needed, for cuda-graph.

[narendasan]

Cool, I think this is a totally reasonable approach to WAR DDS cases. One thing that came to mind is if there is a way to make this validator generic because I think it makes sense to have "different" converters which use a the same internal defensive implementation but may have different restrictions with different codepaths like dynamic shape, dds, etc. to register multiple times and use validators to distinguish the restrictions. So a validator could take an arg index and a type allow list and deny list and pass or fail the converter @zewenli98 @apbose what do you think

Maybe a API similar to enforce_tensors