fix(numpy): aliases of builtin types is deprecated as of numpy 1.20

[mwtoews]

These changes shouldn't change anything, since types like np.int are alias to builtin int types. These alias are deprecated as of numpy 1.20.

A few considerations to note:

• There might be some cases where (e.g.) np.int32 should have been specified. This should only be important while reading/writing unformatted data.
• On Windows, the builtin int for NumPy means np.int32, while on other systems it is np.int64.

[codecov]

Codecov Report

Merging #1052 (b400718) into develop (3ccdd57) will decrease coverage by 16.390%.
The diff coverage is 0.000%.

@@              Coverage Diff              @@
##           develop    #1052        +/-   ##
=============================================
- Coverage   16.390%   0.000%   -16.391%     
=============================================
  Files          225      225                
  Lines        50699    50699                
=============================================
- Hits          8310        0      -8310     
- Misses       42389    50699      +8310     

Impacted Files	Coverage Δ
flopy/discretization/grid.py	0.000% <ø> (-46.875%)	⬇️
flopy/discretization/unstructuredgrid.py	0.000% <0.000%> (-16.863%)	⬇️
flopy/export/shapefile_utils.py	0.000% <0.000%> (-10.968%)	⬇️
flopy/export/utils.py	0.000% <0.000%> (-3.671%)	⬇️
flopy/export/vtk.py	0.000% <0.000%> (-7.340%)	⬇️
flopy/mf6/coordinates/simulationtime.py	0.000% <ø> (-38.889%)	⬇️
flopy/mf6/data/mffileaccess.py	0.000% <0.000%> (-6.043%)	⬇️
flopy/mf6/data/mfstructure.py	0.000% <0.000%> (-13.645%)	⬇️
flopy/mf6/mfmodel.py	0.000% <0.000%> (-13.410%)	⬇️
flopy/mf6/utils/mfobservation.py	0.000% <0.000%> (-12.555%)	⬇️
... and 266 more

[christianlangevin]

Looks like our testing is being bitten by a weird shapely issue (as also mentioned here: #1033). @mwtoews any idea what's going on here? This seems like it's in your wheelhouse.

By the way, thanks for all this cleanup on our np.aliases. Very helpful and good to know.

[christianlangevin]

This PR would have passed, except for the geos version issue, which was fixed by a separate PR.