Use deepdish in data_format.py

It has been pointed out by Greg Ashton that if we use deepdish to load in the BILBY results file, then the posterior samples are no longer separated into separate blocks. He suspects that the reason why this is the case if because the posterior samples are a pandas data frame and h5py does not understand the stored data.

Therefore, in data_format.py, we should load in the BILBY results file with deepdish rather than h5py. The rest of the code in main.py does not need to be changed as that only reads the temp h5 file created from data_format.py which is not a pandas data frame.