Pyroa

Fixed issue with initialising walkers

- Changed the priors in A and B expressed in units of RMS and mean to the MAD and mean. This should better handle lightcurves with lots of outliers when setting up the priors for A and B.

Changes:

- The effective number of parameters when delay_dist=True now only uses the driving lightcurve. This has a small effect but with this change, the number of parameters only depends on the smoothness of the driving lightcurve and not the width of the delay distributions
- Priors on the A and B (scale and shift) parameters are handled differently. The user now specifies unitless lower and upper limits as factors of the calculated RMS (A) and mean (B) of each lightcurve. This allows the prior range to vary for each lightcurve to provide more sensible limits especially in cases where the fluxes of each lightcurve vary drastically.
- The initial values of the walkers are now setup to better sample the parameter space, now set to 0.2*initial_value.
- Fixed an issue where delay_dist=True did not allow negative mean lags.
- Fixed error for the in-line output of the best-fit lags.

- Added new utility functions (Utils.py) to analyse the outputs of PyROA. See the Utils_Tutorial for examples of usage. These include displaying lightcurve(+residuals), Convergence plots, corner and chain plots by parameter, Lag spectrum plot and Flux-Flux analysis plot.

- Changes to the intercalibration files were added. It now outputs, for every datum the original telescope, ROA model+uncertainty at the point, and degree of freedom value.

- Re-implemented the inclusion of a transfer function on the lightcurves with a numerical convolution, allowing any delay distribution function to be used.
- Added a backed using emcee's backend to save and resume progress.

- Fixed bug in calculating the slow component.