Feature request for the implementation of variance, standard deviation, and other statistical MIL Ops

## 🌱 Describe your Feature Request
- The problem is thet sometimes one requires to calculate the variance or standard deviation of tensors in a neural network. But coremltools doesn't provide these right out of the box.
- CoreML / iOS version you are using? IDK about CoreML version but I'm working on iOS 14.5 app.
- I have not contributed here but yes I've contributed to S4TF project but only in Swift. If this project requires me to write in C++ then it won't be possible for me but if you guys are okay with me coding in Python then I'm fine with contributing.

## Use cases
- Some image generative models require these operations. It's not only limited to just these kind of models.
- One example that comes to my mind at the moment is StyleGAN2's generator network requires calculating standard deviation.

## Describe alternatives you've considered
I have considered implementing my own function for calculating variance. My implementation is available in issue #1202's discussion comments.

But then, not _every_ developer will be able to implement these (or other) tensor operations _correctly_. Furthermore, I think these are very obvious statistical operations that a user might expect to be available right out of the box because commonly used frameworks like PyTorch and TensorFlow provide these.

## Additional context
N/A

No files, tests, or entry points are named. Start by reviewing the implementation discussed in issue #1202 and the existing MIL operation structure, then clarify which statistical operations and implementation language are in scope. Done would require the agreed operations to be available for tensor models with validation of their behavior.

Written by the indexing model from the issue text.