refactor: streamline LoRA parameter restoration for JAX/NNX models

[RexBearIU]

Description

Refactors the LoRA parameter restoration utility (restore_lora_from_path) to natively accept and merge parameters of JAX/NNX model wrapper classes.
Standardizing on the NNX state tree representation removes complex state-container unpackings, enhancing trainer safety and maintainability.

Key Changes:

• LoRA Utilities (src/maxtext/utils/lora_utils.py): Overhauled restore_lora_from_path to directly query and merge parameters of NNX wrapper
  classes.
• SFT Trainer (src/maxtext/trainers/post_train/sft/train_sft.py): Simplified parameter setup sequence to cleanly map restored LoRA states
  directly onto the model.

Tests

• Verified via unit testing for LoRA restore functions:

python3 -m pytest tests/post_training/unit/lora_utils_test.py 

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 70.58824% with 10 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/maxtext/utils/lora_utils.py	75.00%	5 Missing and 3 partials ⚠️
src/maxtext/trainers/post_train/sft/train_sft.py	0.00%	2 Missing ⚠️

📢 Thoughts on this report? Let us know!