Improve GAM plots

[damianooldoni]

By default we prefer using the number of occurrences/occupied grid cells at class level as covariate in apply_gam().
The returned GAM plots can look quite weird in this case. The GAM fitting line and the colored dots could be very far from the black dots showing the real values. See graph below.

Some thoughts: if the covariate is used:

• the black dots should not be placed in the same plot of the GAM fitting
• 3 subplots: real values, real values at class level, GAM fitting

This issue has been mentioned in trias-project/indicators#72, trias-project/indicators#106. SOlving it means solving them as well.

[damianooldoni]

@soriadelva: is this something you would like to have solved already for the trend analysis of Wallonian protected areas?

[soriadelva]

@soriadelva: is this something you would like to have solved already for the trend analysis of Wallonian protected areas?

Ideally, yes. If you like, I can have a look at this!