Make Raster.to_pointcloud() modular and add Raster.get_mask() for memory-efficient operations requiring mask

[rhugonnet]

Summary

This PR improves Raster.to_pointcloud() in terms of argument consistency, modularity, and makes its subsampling functionality only returns valid values (while before it returned any random sample) to also be consistent with other subsampling methods.
Additionally, this PR adds a non-public Raster._load_only_mask() function called by a new public function Raster.get_mask(), which allows to get the nodata mask without loading the array into memory. This is used in .to_pointcloud() to perform sampling of only valid values without loading the array in memory 😄.

The PR also adds a lot of tests!

Detailed changes

This PR:

• Renames Raster.to_points() into Raster.to_pointcloud() for consistent naming in preparation for Add PointCloud subclass of Vector #492,
• Define a main data_column_name argument for Raster.to_pointcloud() to associate the point cloud to a certain data_band (instead of having a hard-coded "b1", "b2", etc), and proposes to store all columns using the new argument store_auxiliary_bands (naming to be made consistent with PointCloud class but should be fairly close),
• Adds a random_state argument for subsample,
• Changes the subsampling done to rely on subsample_array(), after calling a new _load_only_mask function to get the valid mask from disk without loading the raster array,
• Modifies _load_rio to allow to return only the mask by calling rasterio.read_masks(),
• Adds a Raster.get_mask() function that returns the raster mask consistently as a boolean array (to circumvent the issue of having a single False value, see Deal consistently with Raster.data.mask=False #505) and without loading the raster array.

Resolves #499
Resolves #504
Resolves #505
Resolves #502

TO-DO:

• Add tests for subsample=1 in subsample_array(),
• Add tests for _load_only_mask,
• Define a function get_mask that works without the array? See Deal consistently with Raster.data.mask=False #505.

[adehecq]

Apart some minor comments, looking good to me. Thanks !!