chore: Refactor to implement model plugin - part 1

[Fei-Guo]

This is the first PR of a series of refactoring changes for moving all model specific configurations into preset directory so that the core controllers will be model agnostic.

• This change introduces Kaito model plugin.
• This change adds the common model interface.
• This change adds corresponding inference implementation for llama-2-chat model family.

NOTE: the added code is not activated (used) in the controller yet.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (e8615bd) 77.05% compared to head (4f8aa4f) 77.05%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #183   +/-   ##
=======================================
  Coverage   77.05%   77.05%           
=======================================
  Files           9        9           
  Lines        1142     1142           
=======================================
  Hits          880      880           
  Misses        226      226           
  Partials       36       36           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.