GitLab

Previously we tokenized and counted tokens to stop when max tokens was reached. Now we let the mistral.rs engine do it which saves the extra tokenization step.

Also dynamo-run prints which engines are compiled in in help message, and some minor lint fixes.