Hanging when generating samples from prior

@sylvia.biscoveanu noticed that PESummary ran for a long time without producing any plots. After some digging we found that PESummary was hanging when trying to generate samples from the prior using bilby. To test this, we ran the following lines of code in a standard python shell:

>>> from bilby.core.result import read_in_result
>>> data = read_in_result("/home/sylvia.biscoveanu/bilby_pipe/runs/151216_psd_marg/median_psd/result/dynesty_IMRPhenomPV2_GW151216_data0_1134293073-165_analysis_H1L1_dynesty_merged_result.json")
>>> data.priors.samples(10)

And data.priors.sample(10) did indeed hang. We let it run for 10 minutes and nothing was produced. After 10 minutes we killed the job and received the following traceback:

>>> data.priors.sample(10)
^CTraceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/charlie.hoy/virtualenvs/pesummary/lib/python3.6/site-packages/bilby/core/prior.py", line 315, in sample
    return self.sample_subset_constrained(keys=list(self.keys()), size=size)
  File "/home/charlie.hoy/virtualenvs/pesummary/lib/python3.6/site-packages/bilby/core/prior.py", line 358, in sample_subset_constrained
    [all_samples[key], samples[key][keep].flatten()])
  File "/home/charlie.hoy/virtualenvs/pesummary/lib/python3.6/site-packages/numpy/core/shape_base.py", line 338, in hstack
    return _nx.concatenate(arrs, 0)
KeyboardInterrupt

I am running bilby==0.5.8. I think @gregory.ashton has also noticed PESummary running slow with a bilby result file and I suspect that it is the same problem.