Checkpoint conversion tool: Optimize to_maxtext & Onboard deepseek2/3/3.2

[shuningjin]

Description

onboard deepseek family

• deepseek2-16b, deepseek3-671b, deepseek3.2-671b

optimize to_maxtext

• use bfloat16 to load and save
• reduce memory by half in all cases
• speedup for large model: e.g., deepseek3-671b, previously impractical (loading alone takes 11hr), now total conversion is 9hr (with 4min loading)
• increase support: can convert model without hf code, as long as there is safetensor (e.g., deepseek3.2)

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.