Fix llama4 test with hf access token

[shuningjin]

Description

Problem: llama4 forward logit check fails with no gated access.

FIXES: b/535231546

Access to model meta-llama/Llama-4-Scout-17B-16E is restricted. You must have access to it and be authenticated to access it. Please log in.

Cause:

• Forward logit check with pre-generated golden logit previously does not required hf access token. However, it becomes required with PR4330, which additionally loads the tokenizer for analysis
• Only "gated model" needs hf access token for this, which includes llama4. Many other modes are not gated and thus remains intact (e.g., deepseek, gpt-oss) in tests.

Solution: pass hf access token in test command

Remark: environment variable will be exported in a separate PR: GoogleCloudPlatform/ml-auto-solutions#1322

Tests

Test on v6e-256

• cmd: https://paste.googleplex.com/5831508681162752
• log: https://cloudlogging.app.goo.gl/keokGV7oK8zLA2vWA

successfully loaded tokenizer

Loading tokenizer from meta-llama/Llama-4-Scout-17B-16E.

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

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!