[contrasting] improve pointwise_gev_study_visualizer.py with a linear regression against the altitude. modify the _maxima_gev depending on the fit method

[contrasting] add a pre-computed of the max_abs value for the shape parameter & for the various moments

[contrasting] add pointwise_gev_study_visualizer.py and add confidence interval for the gev parameters for is_mev fit method

[contrasting] add temporal covariate argument to the margin model class, so that we could use the global temperature as covariate.

[contrasting] display individual model fitted with all data (300-4800). Display only the best non-stationary model if it passes the Anderson fit test.

[exceeding] add -1 to the plot of the uncertainty curve, to show all variables display refer to the same axis. Improve uncertainty histogram plot. add the mean relative difference with the return level

[contrasting] add default pval equal to zero in gnFit (to avoid the program to crash when called by python). refactor anderson test. add anderson test by default for the one fold analysis.

[exceeding] fix code to create a qqplot. (sort only once the empirical quantiles have been computed). Create ALTITUDE_TO_GREY_MASSIF in utils to ensure that the exceedance plots are always done with the same massifs everywhere (and these massifs are only the one that are not grey for the non-stationary gev & gumbel)

[exceeding] update style of the qqplot. clean study_visualizer_for_non_stationary_trends.py, add a mode with only selected models that does not crash.

[exceeding] modify plot settings for the paper. put "imap" instead of "map" to avoid computing 2 times the models

[contrasting] harmonize gumbel altitudinal models. create plot_total_aic.py. remove exception for model with infinite aic.

[contrasting] add quadratic & cross terms Gumbel altitudinal models & fix a recurrent bug that happenning when checking the AIC value

[contrasting] add Gumbel altitudinal models. But so far only the first three models are passing tests (models with quadratic & cross terms are crashing)

This includes models that combine some quadratic aspect & a cross term.
Remove random test

the selection for the model with minimal aic only takes into account models with with finite aic (there are often several models with infinite aic)