Make debug_print more readable

[ricardoV94]

import pytensor
import pytensor.tensor as pt

x = pt.scalar("x")
y = pt.vector("y")
z = x + pt.exp(y)

pytensor.dprint(z)

Elemwise{add,no_inplace} [id A]
 |InplaceDimShuffle{x} [id B]
 | |x [id C]
 |Elemwise{exp,no_inplace} [id D]
   |y [id E]

This contains a lot of detail that is irrelevant for most users:

1. Elemwise{...}. Almost all operations are Elemwise because users don't build graphs with scalars. I suggest we just call Add. The "type" of addition can be guessed from the input types which are visible with print_type=True. This will be even more relevant after Implement graph.vectorize and Blockwise Op #306 where everything that is not an Elemwise will likely be a Blockwise. We can maybe add a kwarg to show the exact Op that is being used.
2. The inplace/no_inplace is seldom important for users. This information is still visible via either print_destroy_map or print_view_map. Maybe we can have a print_memory_map that is equivalent to setting both of those to True.
3. DimShuffles are incredibly common, but very obscure because they are a term only used in PyTensor. I suggest we specialize the name for the following 4 cases:
   1. ExpandDims{axis/axes=dims}
   2. DropDims{axis/axes=dims} (or Squeeze?)
   3. Transpose{axis/axes=dims}
   4. DimShuffle{order} (if it's a mix of the ones above)

The default dprint for the above graph would then look something like:

Add [id A]
 |ExpandDims{axis=0} [id B]
 | |x [id C]
 |Exp [id D]
   |y [id E]

And dprint(print_type=True, print_memory_map=True)

Add [id A] <TensorType(float64, (?,))>
 |ExpandDims{axis=0} [id B] <TensorType(float64, (1,))> v={0: [0]}
 | |x [id C] <TensorType(float64, ())>
 |Exp [id D] <TensorType(float64, (?,))>
   |y [id E] <TensorType(float64, (?,))>

Which I think is way more readable! We can go one step further and rename <TensorType(float64, (?,))> to <Tensor(float64, shape=(?,))> or even <Vector(float64, shape=(?,))>

[twiecki]

This would be great, I have a really hard time parsing everything that's going on in that output usually, but your simplified proposal is much easier to parse.

On Aesara there also was the proposal to use more complex characters like ┕.

[ricardoV94]

On Aesara there also was the proposal to use more complex characters like ┕.

Yeah that would be even nicer!