[refactor] refactor test_quantile_estimator.py. Quantile estimators works for any temporal coordinates.

[paper 1] add plot for the presentation. increase the size of the legend for uncertainty curves. add sample function for the GevParams class.

[paper 1] add plot with the difference in return level (and relative difference) for the big shapes. we conclude that there is big difference indeed

[paper 1] add upper and lower bound into the gev param object. Add script to run fit wihout maximum, and then compare the upper bound with the maximum values that was removed. The test is ok for all altitudes, the upper bound found was always higher than the maximum values

[Confidence Interval] use fix r.ci.fevd.mle_fixed for the Gumbel case in the Python code. fix the R code one more time for the gradient

[PAPER 1] add gumbel type for mle fit. add study_visualizer_for_shape_repartition both for gumbel VS stationary and stationary VS non-stationary

[PAPER 1] improve uncertainty plot. fix eurocode value for the histogram. modify number of bayesian iterations to 20000 for uncertainty plot

[PAPER 1] fix gelman convergence score formula, take the variance instead of the std. improve bayesian display

[PAPER 1] add params_start_fit_bayesian for gelman convergence. add some random init param for the beginning of the MCMC chain.

[PAPER 1] refactor folder names in the paper snow load folder. add mani gelman convergence test. add the possibility to change the number of iterations for the bayesian fevd estimation

[PAPER 1] taking into account psnow when computing TDRL and also when computing return level uncertainties

[PAPER 1] refactor folder. enable in a single script to evaluate trends and return level uncertainty