WIP: EM-Bright infrastructure for OPA

This is the EM-Bright infrastructure for the OPA. Currently, the full EM-Bright infrastructure cannot be implemented because EM-Bright depends on components of rapid_pe, which has to stay in Python2.7 for the time being. So, for the OPA the infrastructure is only going to use the detection pipeline point estimate. Currently, the only classification implemented is the presence of neutron star, with the remnant mass classification to be incorporated as soon as possible (need coordination with Francesco to get his inference code in PyCBC moved into LALinference). At the present moment, there is a placeholder function called source_classification in the em_bright.py code that just accepts the point estimates and returns 0 or 1 depending upon whether the mass2 is greater than or less than 3.0 solar mass respectively. It returns the remnant mass probability as a NaN. When the actual EM-Bright code is moved to LALinference, renamed source_classification.py, this function will disappear and instead, that module will be imported. Due to the high degree of similarity between EM-Bright and BAYESTAR, I have been modeling the gwcelery implementation of EM-Bright based on what already exists in BAYESTAR.