Crash on full infinites values at warmup

[SBouchard01]

If the first sampling of logl values is full of np.inf during the warmup, the code will crash at line 461 as finite_idx will be empty.

The error can be reproduced easily with the following snippet (or by setting a log_likelihood constant at -np.inf):

import numpy as np
logl = np.asarray([np.inf, np.inf, np.inf, np.inf, np.inf, np.inf, np.inf, np.inf, np.inf, np.inf])
x = np.arange(len(logl)) # Technically logl is sampled from x but this example is just here to test logl

inf_logl_mask = np.isinf(logl)
if np.any(inf_logl_mask):
    all_idx = np.arange(len(x))
    infinite_idx = all_idx[inf_logl_mask]
    finite_idx = all_idx[~inf_logl_mask]
    idx = np.random.choice(finite_idx, size=len(infinite_idx), replace=True)

Somehow, calling Sampler.sample_prior(Sampler.n_prior) before run() seems to mitigate the issue a bit (I think this might be because the state of numpy.random changes when the sampling is called?)

The error raised

File "lib/python3.12/site-packages/pocomc/sampler.py", line 461, in run
    idx = np.random.choice(finite_idx, size=len(infinite_idx), replace=True)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "numpy/random/mtrand.pyx", line 970, in numpy.random.mtrand.RandomState.choice
ValueError: 'a' cannot be empty unless no samples are taken

Suggestions of solutions

Either handle the error to clarify the reason of the crash, or re-create a sampling until some values are non-infinite (with a fixed number of tries - then raise an error)