Update gather reduce kernel aligning with tpu inference

[NuojCheng]

Description

This PR

• updates ragged gather reduce kernel to match the curent vllm implementation
• Add cost estimate on kernels for potentially better XLA scheduling
• (tempararily) disable ragged_gather_reduce kernel for ring of attention backward pass

Tests

Please describe how you tested this change, and include any instructions and/or
commands to reproduce.

Checklist

Before submitting this PR, please make sure (put X in square brackets):

• I have performed a self-review of my code. For an optional AI review, add the gemini-review label.
• I have necessary comments in my code, particularly in hard-to-understand areas.
• I have run end-to-end tests tests and provided workload links above if applicable.
• I have made or will make corresponding changes to the doc if needed, including adding new documentation pages to the relevant Table of Contents (toctree directive) as explained in our documentation.

[codecov]

Codecov Report

❌ Patch coverage is 7.40741% with 50 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/maxtext/kernels/ragged/ragged_gather_reduce.py	8.51%	43 Missing ⚠️
src/maxtext/kernels/ragged/ragged_gather.py	0.00%	4 Missing ⚠️
src/maxtext/kernels/ragged/ragged_sort.py	0.00%	3 Missing ⚠️

📢 Thoughts on this report? Let us know!