OGGM features vs independent calculations

[nmaffe]

zmin, zmax, zmed are currently imported from oggm. I think using tandemx could be better ?

[nmaffe]

Also points_df['Slope'] = gl_df['Slope'].item() should be replaced.

At training time one idea is to clip slope xarray with geometry and calculate mean slope.
At inference time one idea is to calculate the mean of the interpolated slope values.

[nmaffe]

replace OGGM's Lmax with lmax_with_covex_hull both at training time and inference time