add completion function

[bachvudinh]

As requested in #4, I created a function called completion to receive prompt input and return the output string. Thanks to @phamson02 for working with me on my first commit and my first time diving into a C++ project.

[namtranase]

Thank the team for the hard work. I read your code and have a few suggestions, please consider:

• We can make a class Model to hold all parameters of the model and make sure it has key functions like completion, chat, batching...
  -> This makes sure that you don't need to re-load the model each time you do inference.