Alias duplication for llama3 in config template

[joshbainbridge]

In the default template the 'llama3' alias is used twice, once for groq and again for ollama. The ollama appears to take priority and masks groq. Should these have unique identifiers?

Also related to the ollama config, it currently targets the 'llama3:70b' model. I'd propose that this is changed to 'llama3', the default 8b model. Most users won't have a 40GB GPU required to practically run 70b, and will more likely have the default model already installed.

Great project by the way, really appreciate all the hard work.

[caarlos0]

refs #300

if no --api is specified, it'll use the first one that matches.