[bug][examples] Fix lcb dataset issue in examples dir

[s-chundi]

Fixes #542

The original deepcoder_train.json is a 5.2 GB JSON array. When datasets.load_dataset("json", ...) reads it, PyArrow's JSON reader sets a block_size based on the file size, but block_size is an int32_t — so any file over ~2 GB overflows and crashes with OverflowError: value too large to convert to int32_t.

Changed lcb_dataset.py to output parquet files directly via Dataset.from_list() instead of writing large JSON arrays. Updated data paths to reference the new .parquet files.

Confirmed by training on the resulting parquet files

---

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 2 additional findings.

[gemini-code-assist]

Warning

Gemini is experiencing higher than usual traffic and was unable to create the review. Please try again in a few hours by commenting /gemini review.

[s-chundi]

/gemini review

[gemini-code-assist]

Code Review

This pull request resolves a crash when loading large JSON datasets by switching to the more efficient Parquet format. The changes in lcb_dataset.py correctly replace pandas with datasets for this purpose, and run_lcb.sh is updated accordingly. I've included a suggestion to remove a redundant file creation, which will make the data processing script more efficient.