Goal:
The Checkpointer class introduces a unified approach to our various checkpointing techniques. A user can pass in their checkpointing style (full_state or hf_format), and the checkpointer, via checkpointer.checkpoint, will save the model using the selected method and other techniques (LoRA).

https://github.com/instructlab/training/pull/456

Acceptance Criteria:
Checkpointer class is introduced to the library which allows users to create an object that represents one of the checkpointing methodologies we support. a user should be able to use this object to run training and save checkpoints