chore: update base model to NVIDIA-Nemotron-3-Nano-4B-BF16 across configurations and documentation; adjust max sequence length to 2048; upgrade GitHub Actions cache to v5

build: use python 3.10 and torch 2.11

chore: update pydantic and pydantic-core versions, adjust Python version in README, and enhance CI workflow with caching

Update setup-uv action to version 8.2.0

ci: add GitHub Actions workflow