Multiples bug fixes and add on_train_epoch_start callback

[Edresson]

What it does?

1. Solve KeyError: 'avg_loss_1' error when start_with_eval=True and target_loss is settled. This issue happens because the training keep_avg_target.avg_values is an empty dictionary. It is related to [Bug] KeyError: 'avg_loss_1' crash when training model TTS#2862. This error also happens in training when we try to save a checkpoint before we have updated self.keep_avg_train or self.keep_avg_eval. To solve it this PR also make _pick_target_avg_loss safe and it avoid issues like multiband_melgan Vocoder Fails on Step 10000 With KeyError: 'avg_loss_0' TTS#1608 to happens, if the keep_avg_target.avg_values is empty it will return None and all will be good.
2. It also raises an error if multiple-optimizer setup with grad accumulation and without a custom optimize method. It avoids the user training the model with our implementation that has dangling gradients in multiple-optimizer setup with grad accumulation (I already did it accidently, It is really bad because we can lose training time).
3. It added on_train_epoch_start and on_train_epoch_end callbacks. Currently, the only way to put modules in eval mode model during the training is via on_train_step_start callback, that is called each train_step. It is really slow. Adding this new callback we can do it only one time per epoch. It should decrease the step time for XTTS GPT and XTTS decoder training.