Expanded model back-ends

[sempervictus]

Is your feature request related to a problem? Please describe.
There are currently far too many models which cannot be loaded and run by LocalAI and performance/compatibility optimizations available by adding forks.

Describe the solution you'd like
Review and potentially add runners for projects like:

1. https://github.com/ikawrakow/ik_llama.cpp
2. https://github.com/EricLBuehler/mistral.rs
3. https://github.com/evilsocket/cake
4. ...
   to homologate execution models for system+model nuances under the cover of "front-end."

[ruribe17]

How can i code an external backend for integrating ik_llama.cpp on localai? Could the same backend.proto on this project be use?