Use synthetic data for scripts under src/maxtext/configs/tpu

[aireenmei]

Description

Currently, the dataset type used by scripts under src/maxtext/configs/tpu falls into this pattern:

• Category A (Newer scripts, v6e/, v5p/, v5e/{128b,gpt3_175b}.sh): already use dataset_type=synthetic, but the script has legacy docstring stating the $DATASET_PATH is required, but is actually not used in the command.
• Category B (v5e/{16b,32b,64b}.sh, v5p/{32b,64b,128b,256b,512b,1024b}.sh): has both dataset_type=synthetic and dataset_path=$DATASET_PATH in the command, and dataset_path=$DATASET_PATH is a no-op. Also has the legacy docstring.
• Category C (older scripts v4/*, v5e/{llama2_7b,llama2_13b,llama2_70b}.sh): use the default dataset_type=tfds and requires user to provide DATASET_PATH.

This PR:

• Migrate category C scripts to use dataset_type=synthetic, which aligns with the new convention and allows user to do quick test without dataset preparation.
• Remove the redundant dataset_path=$DATASET_PATH for category B
• Remove legacy docstring for category A and B

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.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!