[Deepseek V4] Add caching support and verify decoding

[Rohan-Bierneni]

Description

PR add caching support to attention_compressed.py and verifies v4 has decoding support with the maxengine/jetstream backend

If the change fixes a bug or a Github issue, please include a link, e.g.,:
FIXES: b/123456

Tests

Synthetic data cmd: https://paste.googleplex.com/5435406769324032
Output: https://paste.googleplex.com/6295276670287872

Decode.py should pass (In progress)

cmd:
output:

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 65.45455% with 57 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/maxtext/layers/attention_compressed.py	65.45%	39 Missing and 18 partials ⚠️

📢 Thoughts on this report? Let us know!

[Rohan-Bierneni]

Created separate pr and merging there: #4444