Adds optional GPU VRAM fit check via --gpu flag.

Users can now compare estimated VRAM requirements with available GPU
memory directly from the CLI.

Example:
llm-plan meta-llama/Meta-Llama-3-8B --gpu 24