BUG: inhomogeneous rescaling with Categorical Priors

[Hen42rik]

Adding a check for np.int64 type which is returned by WeightedCategorical priors.

[Hen42rik]

Here is a MWE to reproduce the bug:

import numpy as np
from bilby.core.prior import ConditionalPriorDict, WeightedCategorical, Gaussian
from bilby.core.likelihood import  Likelihood
from bilby.core.sampler import  run_sampler

class DummyLikelihood(Likelihood):
    def log_likelihood(self, parameters):
        return -0.5 * np.sum([v for v in parameters.values()])**2
    
priordict = ConditionalPriorDict()
priordict["x"] = WeightedCategorical(name="x", ncategories=100)
priordict["y"] = Gaussian(name="y", mu=0, sigma=1)

run_sampler(likelihood=DummyLikelihood(priordict),
            priors=priordict,
            sampler="dynesty")

[ColmTalbot]

I'm happy to get this in as a quick fix, but this behaviour should be fixed more generically in a future release.