catch and improve dynesty loglikelihood errors
When there's a likelihood error, we've never been able to reproduce the failure by calling the wf directly (either just through lalsim or through bilby). We suspect this may be because the print precision of the dumped parameters list is insufficient. Also it would be helpful if it were a dict instead of a list. The error message is straight from dynesty, but @gregory.ashton thinks "a simple try/except around the run_nested could do it.".
My use case is parallel_bilby, but I'm opening the issue here since I assume it's calling dynesty by way of the main bilby package, or at least the same workaround would go into both...?