Exclude RNG state when updating vllm with checkpoint for pre-rl evaluation

[SurbhiJainUSC]

Description

Running RL with Qwen3-0.6B on v5p-8 throws error when num_test_batches>0: https://paste.googleplex.com/6146006073868288

Tests

Tested with Qwen3-0.6B on v5p-8

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.