Change default behaviour of crop to inplace=False

[atedstone]

This PR creates breaking changes to the API

Resolves part of #432: Raster.crop() is now inplace=False by default. This aligns its functionality with Raster.reproject() and with other projects more generally such as xarray.

The PR also fixes a bug in test_projtools:test_utm_to_epsg. A KeyError was being returned by proj_tools.utm_to_epsg():

        with pytest.raises(TypeError):
            pt.utm_to_epsg({"utm": "10N"})  # type: ignore

This is now fixed by requiring a str type in utm_to_epsg(), allowing all tests to pass locally on my machine.

[atedstone]

I should also note that I checked for other possible uses of inplace functionality. As far as I see, only crop() has this option, so there are no further functions to change.

[rhugonnet]

Great @atedstone!
Curious that only these tests broke... I think by leaving the None return for inplace there might be more to change! (in the doc examples in particular)

[atedstone]

I now realise that Vector.crop() also needs changing - will do this and add to this PR.

[atedstone]

On my side this PR is ready for review and merging.

[rhugonnet]

Amazing!!
So it's having inplace: bool = ... repeated in an @overload above the main function definition that made MyPy happy in the end?

[rhugonnet]

This should still return None if it is inplace and crop_img was already written to .data, no? 🤔

[adehecq]

I agree, we should not return self, if the data is modified in place, otherwise this becomes confusing...

[atedstone]

hmm, ok it looks like this snuck through, which was not the intention. I did this when having problems with @overload but that got resolved. Will sort soon, for the moment I'll open a bug