Add param_group name for BaseFinetuningCallback

Description & Motivation

Add option to pass param_group name when we use unfreeze_and_add_param_group method in finetuning callback for better LR logging.
As it said in documentation

optimizer has multiple parameter groups they will be named Adam/pg1, Adam/pg2

so when we add new param_group, it will be shown as adam/pg2

Pitch

Change from:
optimizer.add_param_group({"params": params, "lr": params_lr / denom_lr})
to
optimizer.add_param_group({"params": params, "lr": params_lr / denom_lr, "name": unfreeze_params_name})

Alternatives

No response

Additional context

No response

cc @lantiga @borda

Search the finetuning callback implementation for the unfreeze_and_add_param_group method and inspect how it calls optimizer.add_param_group. Pass the unfreeze parameter-group name through to the optimizer so the newly added group uses that name in learning-rate logging. Verify that the resulting logs identify the group by its supplied name.