Add VLA fine-tuning example

[shorbaji]

Add an example for Vision-Language-Action (VLA) model fine-tuning using Ray Train on Anyscale.