Th228 fits: pvalue, fwhm uncertainty, iterative fits (optional)

[LisaSchlueter]

1. gof.jl contains 3 functions to calculate the p-value as a goodness-of-fit for the calibration peak fits.

• p_value (default). based on least-squares
• p_value_LogLikeRatio based on likelihood ratio
• p_value_LogLikeRatio Monte Carlo to sample loglikelihood

2. iterative_fits as a new argument for fit_peaks. default is false. If true and the fit parameter covariance matrix is not semi positive definite, the peak fit is repeated with low_e_tail=false
3. nearestSPD(): finds nearest semi positive definite matrix
4. uncertainties on fwhm now with correlated sampling. see get_mc_value_shapes