Get rid of floatX calls inside distributions

[ricardoV94]

Description

These calls make inputs that are valid tensor_like fail:

import pymc as pm

with pm.Model() as m:
    x = pm.Normal("x")
    y = pm.Beta("y", [x+1, x-1], 1)  # Fails
    y = pm.Beta("y", pm.math.stack([x+1, x-1]), 1)  # works

TypeError: float() argument must be a string or a real number, not 'TensorVariable'
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
  File "/home/ricardo/.conda/envs/pymc-experimental/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3505, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-8-9eb4432272c4>", line 5, in <module>
    y = pm.Beta("y", [x+1, x-1], 1)
  File "/home/ricardo/.conda/envs/pymc-experimental/lib/python3.10/site-packages/pymc/distributions/distribution.py", line 310, in __new__
    rv_out = cls.dist(*args, **kwargs)
  File "/home/ricardo/.conda/envs/pymc-experimental/lib/python3.10/site-packages/pymc/distributions/continuous.py", line 1142, in dist
    alpha = pt.as_tensor_variable(floatX(alpha))
  File "/home/ricardo/.conda/envs/pymc-experimental/lib/python3.10/site-packages/pymc/pytensorf.py", line 439, in floatX
    return np.asarray(X, dtype=pytensor.config.floatX)
ValueError: setting an array element with a sequence.

PyTensor has a builtin mechanism to control the promotion of inputs to float that should be accessible to users.

[jaharvey8]

I could help out here. But to make sure I understand what's needed, each distribution has a bunch of

"as_tensor_variable(floatX(something))"

which needs to be changed to

as_tensor_variable(something)

correct?

[ricardoV94]

Yes, it may break some tests but we can assess on a case by case once that happens.