[ET-VK][ops] Extend arange, clamp, and index.Tensor support

[SS-JIA]

Stack from ghstack (oldest at bottom):

• -> [ET-VK][ops] Extend arange, clamp, and index.Tensor support #22481

Support runtime-varying arange bounds, fractional and negative steps, and symbolic clamp bounds for integer and floating-point tensors. Extend buffer index.Tensor to gather on any axis while preserving multidimensional index shapes.

Authored with Codex.

Differential Revision: D118507314

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/22481

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 8d67228 with merge base 902fcf5 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

[msluszniak]

This is a better solution that these two: #22406 and #22399 which are my workarounds rather than direct solutions. Please double check if this PR covers them completely.

[msluszniak]

Checked, and it covers both, so I have acted on it:

• Vulkan: support index.Tensor with the index on any dimension #22406 (index.Tensor on any axis) is closed in favour of this. One thing it does not carry over is filed as [ET-VK] TagMemoryMetaPass skips a tensor that sits in a list alongside None #22510: a tensor sitting in a list alongside None is skipped by TagMemoryMetaPass, which this sidesteps by forcing CONTIGUOUS_BUFFER rather than closing.
• [ET-VK] Do not partition constant_pad_nd with a symbolic pad #22399 kept only constant_pad_nd, whose pad list is a VALUELIST that Pad.cpp reads with get_int_list(). add_dynamic_clamp_node supersedes its clamp and hardtanh guards, so those are dropped; keeping them would decline nodes this makes work. hardshrink and leaky_relu still read bounds at build time via get_val_or_inf(), but a symbolic lambda or negative_slope is not something a real model produces, so I left them alone. Its batch norm guard moved to [ET-VK] Do not partition batch norm on a non-4d input #22509.

#22399 merges cleanly with this branch.