WIP: Implement Relative Binning in Bilby

Created by: adivijaykumar

Things to Fix:

• Function to generate the waveform for unevenly space frequency array (This should be pretty easy; basically one needs to define a _base waveform in source.py and use lalsim_SimInspiralChooseFDWaveformSequence instead of lalsim_SimInspiralChooseFDWaveform)

• Add new frequency_domain_source_model in source.py. This source model should return the fiducial waveform (and the summary data) using the full frequency array and the likelihood calculation waveforms using the binned array.

• Makes changes to compute_relative_ratio

• For summary data calculation (in compute_as_and_bs), use noise_weighted_inner_product

• Remove num_points from setup_bins.

• Issues with zeros of the waveform (this is probably occuring in cases where the waveform is getting terminated before fmax)

• The calculation is being done only for parameter_dictionary and not for self.parameters. Correct this. (#4 - @Kruthi24)

• Fix and check get_detector_response_relative_binning

• Remove log_likelihood_ratio_approx (#5 - @Kruthi24)

• Implement scipy maximization (@Kruthi24)

• See if new summary_data dictionary is the right choice.

• Add unit tests.

• Implement time, distance and phase marginalization.

• Remove all print statements and replace with logger instances. Also remove all instances of sys.

• Write documentation

• Clean up comments

• Remove imports of matplotlib

...and more to be added in due time.