PyTensor baby steps example focuses too much on dtypes

### Description

https://pytensor.readthedocs.io/en/latest/tutorial/adding.html

I think we should leave types discussion for later and emphasize the laziness until compilation instead.

Could also mention the general `tensor` class