Add caching for GatedDeltaNetCache (linear attention models)

[Rohan-Bierneni]

This pr adds caching support to the Gated Delta Net implementation for Qwen3 Next. This is the first non-standard attention caching support in MaxText and needed changes to both MaxEngine code as well as qwen3 next implementation code.

If the change fixes a bug or a Github issue, please include a link, e.g.,:
FIXES: b/448407748

Tests

Ran the MaxText.decode command and got meaningful output: https://paste.googleplex.com/5431522368815104

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 38.09524% with 39 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/MaxText/layers/decoders.py	9.52%	17 Missing and 2 partials ⚠️
src/MaxText/layers/qwen3.py	60.00%	9 Missing and 5 partials ⚠️
src/MaxText/maxengine.py	14.28%	6 Missing ⚠️

📢 Thoughts on this report? Let us know!