Move to python 3.10 in CI and remove python 3.7

[rhugonnet]

Main change in the title.

Also ran into the issue of RuntimeWarning with NaNs of #260 again. I fixed it by using np.full(shape, np.nan) to initialize a NaN array instead of multiplying or adding np.zeros(), np.ones() or np.empty() with np.nan like we have the (now bad?) habit of doing.
The dtype of np.full (previously np.empty) is not set to dem.dtype anymore. This change was triggered by failing doctests with this dtype forcing because, if a DEM is an integer, terrain attributes derived from it would be forced to integers (and some attributes are between 0 and 1). The failure of tests only happened in this PR because the dtype argument works differently if initiated in np.full or if by np.empy later multiplied by NaN. The np.empty multiplied by NaN actually ignored the dtype all along.

Added a point on this in the wiki (https://github.com/GlacioHack/xdem/wiki/Stuck-on-a-mysterious-error-during-CI%3F):

4. The unexpected: Initialization of np.ndarray full of NaNs could historically be done in many ways: multiplying or summing a np.zeros, np.ones or np.empty with a np.nan. However, recently, RuntimeWarning have started to pop up a bit randomly with these methods, resulting in test failres. We could not track down the exact version change that triggered this (might be unexpected behaviour, so not in the logs), but it does seems to depend on the version of numpy. Solution: use np.full(shape, np.nan, dtype) to initialize those arrays!

values = np.empty((n_bins, unique_indices.shape[0]), dtype=float) * np.nan
E       RuntimeWarning: invalid value encountered in multiply

Resolves #249
Resolves #260

[adehecq]

Thanks !