Remove ds_config scheuduler params to prevent deepseed from creating scheduler for ref_model

[Ben-Schneider-code]

What does this PR do?

Fixes #2154

My opinion is that popping the scheduler params off the deepcopied ds_config is the best way to solve this. This prevents deepspeed from attempting to build a scheduler for the engine (which is fine because any returned scheduler is ignored anyways). I explored some other avenues like using the Deepspeed inference engine instead but the different config would make it a nightmare.

Other:

• Removed redundant None check, as ref_model is None checked before _prepare_deepspeed is called:

if self.ref_model is None:
    if not (self.is_peft_model or self.precompute_ref_log_probs):
        raise ValueError(
            "No reference model and model is not a Peft model. Try setting `precompute_ref_log_probs=True`"
        )
    if args.sync_ref_model:
        raise ValueError(
            "You currently cannot use `ref_model=None` with TR-DPO method. Please provide `ref_model`."
        )
else:
    if self.is_deepspeed_enabled:
        self.ref_model = self._prepare_deepspeed(self.ref_model)
    else:
        self.ref_model = self.accelerator.prepare_model(self.ref_model, evaluation_mode=True)

also even if _prepare_deepspeed ends up getting used elsewhere passing model=None to deepspeed will hit an assert anyways.

• Added a comment to clarfify that _prepare_deepspeed can only be used on reference models that won't be trained
• Fixed typo

@qgallouedec sorry for the delay on such a simple fix.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a GitHub issue? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines.
• Did you write any new necessary tests?