PIT with nn.CrossEntropyLoss() from MIRNet

## 🚀 Feature
This is to assign speaker ID of each separated signals with PIT.

### Motivation

Authors from [MIRNet](http://www.interspeech2020.org/uploadfile/pdf/Thu-2-7-3.pdf) considers all of the loss terms by calculating every possible permutation of candidate pairs. They computed the PIT loss on estimated speaker identity information. In addition, they use cross-entropy loss with a classifier for the speaker embeddings.

### What you'd like

The entire training criterion is as follows:

The issue provides no repository files, tests, or entry points. Start by reading the linked MIRNet paper and the requested training-criterion image, then locate the existing loss and PIT-related entry points in the repository. Done means the requested speaker-identity PIT criterion with cross-entropy is supported and covered by validation against the described formulation.

Written by the indexing model from the issue text.