Cannot save editable installs with hdf5 without conda

When there is no conda available, we take version information from pip. If there are editable installed packages, this version information includes an editable_package_location flag for some packages and NaN for all others.

This is then not writeable as the first element is usually np.nan and so falls into the wrong branch of this logic.

When this happens Bilby dumps the result to a pickle file.

Here is an example of the message that shows.

19:43 bilby ERROR   : 

Saving the data has failed with the following message:
No conversion path for dtype: dtype('<U32')
Data has been dumped to result/label_result.pkl.