Primary issues that this resolves:
- Include many different variables in the PE rather than a small subset
- Use only 100s to estimate the PSD
- Use an alpha=0.25 Tukey window (matched to data generated by
  lalinference)
Closes #23

Moritz Huebner: Fixed with new prior class and removed ra, dec, ... parameters from the source function definitions.

Unify parameter and prior classes

See merge request Monash/peyote!17

move around the detector response to the signal into a new function

See merge request Monash/peyote!15

# Conflicts:
#	peyote/detector.py
#	peyote/likelihood.py
#	tutorials/BasicTutorial.py
#	tutorials/standard_data.txt

Moritz Huebner: Some simplifications in the initialise_parameters() method and turned around how keys are being checked in verify_priors