Replace training loop in peft_prompt_tuning with HF Trainer

[gkumbhat]

Description

Currently we use a custom training loop along with HF accelerator for training. This story is to replace that with HF Trainer similar to what we implemented for fine-tuning.

Tasks

1. Implement HF Trainer in peft prompt tuning module
2. Make sure all unit tests pass and add any unit tests necessary
3. Evaluate newly implemented training loop with RTE and CNN datasets using:
   4. Bloom-560m
   5. t5-base

Acceptance Criteria

• Unit tests cover new/changed code
• Example modified to expose any new parameter we may need

[chakrn]

Moving to In Review (see linked PR)

[gkumbhat]

Testing done for this PR merge: #203 (comment)