Parameterize bos_id in inference_microbenchmark.py

[A9isha]

Description

Problem:
inference_microbenchmark.py crashes when running models without a BOS token (like Qwen3) because it hardcodes is_bos=True during tokenization. This causes JetStream's HuggingFaceTokenizer to prepend None to the token list, resulting in a JAX TypeError.

Solution:
Dynamically set is_bos = tokenizer_model.bos_id is not None to avoid prepending None for models that don't have a BOS token.

FIXES: b/508334388

Tests

Ran the following command successfully on a v5p-8

python3 -m maxtext.inference.inference_microbenchmark   model_name=qwen3-4b   tokenizer_path=Qwen/Qwen3-4B   per_device_batch_size=1   max_prefill_predict_length=1024   max_target_length=2048   ici_autoregressive_parallelism=4   scan_layers=false   weight_dtype=bfloat16   attention=dot_product

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.

[parambole]

LGTM 🔥