[projection snowfall] modify altitude_class to altitude_group in visualizer_for_projection_ensemble.py. fix index for ensemble dataset.  add plot_relative_change_in_return_level.py

[projection snowfall] add functions_utils.py. modify one_fold_fit.py so that it may support one altitude fit.

[projection snowfall] refactor nb_params. so that is it associated to the margin_function rather than to the margin_model.

[projection snowfall] adapt the one_fold_fit.py code to handle fit with a single altitude. Modify altitude_class by altitude_group, adapt tests. Modify the DefaultAltitudeGroup. adapt main_projections_ensemble.py

[projection snowfall] add spline simulation. implement splines for k=1. pass test_gev_temporal_spline.py

[projection snowfall] fix one issue for evgam coefficients. refactor param_name_to_dim -> param_name_to_dims in the result objects.

[projection snowfall] add spline_margin_function.py and test with it. test_gev_temporal_spline.py is not yet validated for the nllh.

[projection snowfall] add evgam call for abstract_temporal_linear_margin_model.py. account for log_scale in the margin_function. rename get_coord_df to get_r_dataframe_from_python_dataframe. create evgam_fixed.R for debugging

[projection snowfall] remove test with respect to fevd bayesian, due to differences with the test. i do not use it, so it does not matter. But maybe with the change of R version, the code changed. still try to install evgam

[projections] add visualizer_for_sensitivity.py based on visualizer_for_projection_ensemble.py. add temperature_to_year.py file.
Fix max abs recurrent issue (when it was equal to zero). add one test for one fold fit, that was that due to a scale parameter that was close to zero, and was causing instabilities in the nllh value. This is fixed by changing the check for implausible in one fold fit, and by saying that if the scale parameter is close to zero then the parameter is undefined.

change alpha confidence uncertainty to 80% uncertainty interval when i run the main code.
create 3 qqplots overlaid, one for each altitude.
remove the exclusion of some bootstrap samples (in particular we remove the threshold equal to 2000)

[contrasting] add plot to compare the size of confidence interval between the pointwise approach and the piecewise elevational-temporal approach. bootstrap level is set to 100

[contrasting] implement bootstrap confidence interval both for pointwise stationary models and for elevation-temporal models. accelerate bootstraping with multiprocessing & storing of fitted functions only (storing the estimator was impossible).

[contrasting] improve plots. add comma to the name of the models. change altitude to elevation in several plots.

[contrasting] for preliminary_analysis.py from now we use fevd fit method & fit on series with more than 90% of non zero maxima. improve coherence plot

[contrasting] add general test for test_gev_spatio_temporal_polynomial_extremes_mle.py. refactor a bit

[contrasting]  fix test_gev_spatio_temporal_polynomial_extremes_mle.py for computing the confidence interval. improve coherence curve. remove sensitivity checks in one fold fit. add significant percentages for histograms of models selected. deactivate some tests for speed purpose

[contrasting] add xlabel, remove 300m from the lower altitude group. add condition that we consider only altitude for which the massif has less than 10% of non zero annual maxima. return period is changed to 100.

[contrasting] consider also the stationary model for selection. consider a level of altitude only if at least 2 altitudes are present & the reference altitude is present.

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