Validation for Lorentzian

Tests need to be added to ensure that applying a Lorentzian resolution works as intended, as is done for the Gaussian.
The Gaussian distribution uses a similar test from nMOLDYN as a benchmark for comparison, though as far as I'm aware nMOLDYN does not have a Lorentzian option.

Implementation ideas:
- create a resolution file that's exactly the Lorentzian and use that
- find another program to create a benchmark data file from for a more independent benchmark. (e.g. Mantid)
- create a Lorentzian SQw Observable and directly compare it to the one the code creates.
- Create two Control objects which are the same except one is passed with a Lorentzian resolution of certain FWHM, and the other is passed with no resolution. The one with no resolution then manually has its SQw convolved with the Lorentzian function, and then both are refined for comparison
- Look at whether the results with various FWHMs are within sensible limits
- Take an existing data file, and by hand 'reverse-engineer' noise back into it (i.e. add a Lorentzian) then see if MDMC gets back to the original data successfully.