support for `__len__`

Len works on numpy

from numpy import array

vector = array([1, 2, 3])
len(vector)  # 3

matrix = array([    
    [1, 2], 
    [3, 2], 
    [5, 4],
])
len(matrix)  # 3

But not with pt.as_tensor_variable

from pytensor.tensor import as_tensor_variable as array

# Same as above

which results in:

[ins] In [100,000,000,000,000]: matrix
Out[100,000,000,000,000]: TensorConstant(TensorType(int64, shape=(3, 2)), data=array([[1, ... [5, 4]]))

[ins] In [100,000,000,000,001]: len(matrix)
----------------------------------------------------------------
TypeError                      Traceback (most recent call last)
Cell In[100,000,000,000,001], line 1
----> 1 len(matrix)

TypeError: object of type 'TensorConstant' has no len()

My usecase

import numpy as np

import pytensor.tensor as pt
import pymc as pm


def build_model(y) -> pm.Model:
    coords = {
        "idx": np.arange(len(y))
    }
    model = pm.Model(coords=coords)
    with model: 
        ...

    return model


model = build_model([1, 2, 3])
model = build_model(np.array([1, 2, 3]))
# Failure as above
model = build_model(pt.as_tensor_variable([1, 2, 3]))

Start at the pt.as_tensor_variable entry point and inspect how the resulting TensorConstant handles Python's len() operation. Use the issue's vector and matrix examples as checks, including the PyMC coords use case; done means len() works for tensor variables as it does for NumPy arrays.