Subscript-crop Raster or Vector by bracket call [], consistency and bug fixes

[rhugonnet]

Summary

This PR updates the get_geoprojected_bounds function, moving its core to projtools to make it consistent and accessible for both Raster and Vector classes.
Then, it adds the same options in Vector.crop() than for Raster.crop() (= passing another Raster, Vector or a list of bounds).
Finally, it implements the __getitem__ (= []) method for Raster and Vector, which calls their underlying crop functionality (= subsetting), as suggested previously by @erikmannerfelt.

rst = Raster(fn_rst)
vec = Vector(fn_vec)

# Get only Raster on extent of Vector
rst = rst[vec]

Other than that, several bug fixes and test additions.

Discussion on the __getitem__ implementation

Several things:

• We could add the option of provides slice objects, which would then be used on self.data for Raster (subsetting np.MaskedArray in 2D) and self.ds on Vector (subsetting gpd.GeoDataFrame in 1D).
• We could, by default, reproject instead of crop when it is Raster[Raster] (only one where it makes sense)? The resampling algorithm could be passed as second argument Raster[Raster, "bilinear"].

Related issues and more details

This PR:

• Resolves Remove warning when using Vector.create_mask() only with xres argument #276,
• Resolves Shapely triggers DeprecationWarnings #268,
• Resolves Raster.crop does not reproject the bounding box of a Raster object provided as cropGeom #262,
• Resolves Subscript a Raster with a Vector #213,
• Resolves Polygonize does not support all Raster dtypes #333.

Also it:

• Removes the dependency of projtools.py on raster.py and geovector.py,
• Moves the core of the get_geoprojected_bounds function in Raster to projtools.py,
• Adds get_geoprojected_bounds as a method for the Vector class,
• Mirrors the crop functionality of the Raster class for Vector objects (can provide any type of geometry: bounds, Raster, Vector),
• Adds more test for Raster.polygonize(),
• Solves a new bug with checking void in intersection triggered by package updates.

[erikmannerfelt]

Hey Romain!
These additions are awesome. You're really contributing to this being maintainable in the long term! On the subject of maintainability, I have some comments on details that I struggle to understand, but generally, it's great!

Regarding your two suggestions about the subscripting:

• On the first, do you mean to slice by indices like arr.isel() or df.iloc[] in xarray and pandas, respectively? If so, it would actually be pretty cool, but I struggle to see a use case. For iterating along blocks perhaps? Or subsetting for debug purposes? Those two are good arguments in my head, but for the amount of complexity they introduce, it might not be a great idea unless there are more uses.
• I personally think the subscripting syntax should only use crop. Reproject can be very cpu-intensive and should in my opinion always be explicit. Cropping, however, is really fast, so there the subscripting syntax makes a ton of sense to me.

[adehecq]

Great idea to make crop more consistent and thanks for fixing all the small issues.
The brackets are a nice feature, but they would need to be documented at some point, otherwise, we'll just forget that it's possible !

[rhugonnet]

Thanks a lot! All updated, merging 😄