ci(diffusion): remove local_dir and post process directly on cache

[thomasdhc]

What does this PR do ?

• diffusion_finetune_launcher.sh was calling snapshot_download(..., local_dir=$DATA_DIR/raw), which bypasses HF_HUB_CACHE (modernhuggingface_hub writes only to local_dir). The dataset never landed in the shared cache, so offline reruns failed with LocalEntryNotFoundError.
• Drop local_dir, capture the resolved snapshot path, and point preprocessing at it directly. Cache fills once; subsequent offline runs hit it.

Changelog

• Add specific line by line info of high level changes in this PR.

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines
• Did you write any new necessary tests?
• Did you add or update any necessary documentation?

If you haven't finished some of the above items you can still open "Draft" PR.

Additional Information

• Related to # (issue)