Missing numba dispatch for gammainc_grad

[aseyboldt]

Description

There still seems to be an issue with the gradient of gammainc in numba mode:

import pytensor
import pytensor.tensor as pt

x = pt.vector("x")
y = pt.math.gammainc(x, x)
dx = pytensor.grad(y.sum(), x)
func = pytensor.function([x], dx, mode="NUMBA")
func([1.])

This raises warnings about object mode for gammainc_grad_a and gammainc_grad_b. Those look like ScalarLoop Ops, and I thought we added those recently? Maybe something in the specific ops is raising an error?

original report in nutpie:
pymc-devs/nutpie#163 (comment)

[ricardoV94]

No ScalarLoop is still missing a numba impl, we just fix it failing to even get to obj mode

[ricardoV94]

#299