Migrate PlaceHolderDataIterator from tf to np

[aireenmei]

Description

This is part of the plan of removing TF dependency in MaxText.
Also fix a bug caused by expansion_factor_real_data is type float but some places expect int.

Tests

• PlaceHolderDataIterator is only used when expansion_factor_real_data > 1, on the hosts that are NOT loading real data. The data produced by PlaceHolderDataIterator are all -1 and will be discarded.
  Tested with the following script, set max_checkify=true to error out if training batch contains -1 batch produced by PlaceHolderDataIterator

 python3 /mnt/disks/pd2/xpk/xpk.py workload create \
  --project ${PROJECT} \
  --zone ${ZONE} \
  --cluster ${CLUSTER_NAME} \
  --workload ${RUN_NAME} \
  --base-docker-image ${BASE_IMAGE} \
  --tpu-type ${TPU_TYPE} \
  --num-slices 1 \
  --command "pip install --no-deps -e . && python3 -m MaxText.train src/maxtext/configs/base.yml \
  run_name=${RUN_NAME} base_output_directory=${OUTPUT_DIR} \
  dataset_type=grain dataset_path=gs://maxtext-dataset \
  grain_train_files=gs://maxtext-dataset/array-record/c4/en/3.0.1/c4-train* \
  expansion_factor_real_data=2 max_checkify=true \
  steps=10 enable_checkpointing=false"

cloud log

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 0% with 9 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...axtext/input_pipeline/synthetic_data_processing.py	0.00%	7 Missing ⚠️
...rc/maxtext/input_pipeline/grain_data_processing.py	0.00%	1 Missing ⚠️
...rc/maxtext/input_pipeline/multihost_dataloading.py	0.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[github-actions]

🤖 Hi @aireenmei, I've received your request, and I'm working on it now! You can track my progress in the logs for more details.

[github-actions]

📋 Review Summary

This pull request effectively migrates the synthetic data generator from TensorFlow to NumPy, successfully removing a key dependency. The implementation is clean, straightforward, and achieves the stated goal.

🔍 General Feedback

• The change is well-executed and improves the project by decoupling it from TensorFlow for this component.
• The new implementation using a Python generator is a simple and efficient solution for providing placeholder data.