[hotfix] Use latest VTK version for conda tests

[banesullivan]

The conda tests were using VTK 8.2... why? This updates conda to use whatever the latest version is which seems most appropriate to me. Other versions of VTK are checked in other CI routines.

Either way, this needs to be updated because VTK8.2.2 is incompatible with the latest NumPy 1.24

@pyvista/developers, I plan on merging this quickly to fix broken CI

[codecov]

Codecov Report

Merging #3736 (77da0ca) into main (4bc2064) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main    #3736   +/-   ##
=======================================
  Coverage   94.04%   94.04%           
=======================================
  Files          83       83           
  Lines       18643    18643           
=======================================
  Hits        17532    17532           
  Misses       1111     1111           

[banesullivan]

This cannot pass until #3737 merges

[akaszynski]

We were testing conda using vtk==8.2 because it was the only way we could actually test that 8.2 works. It's a hard version to support because it was never released on PyPI, but I think at this point we should just drop it and (soon) enforce a hard requirement to use vtk>=9.0.

[banesullivan]

should just drop it and (soon) enforce a hard requirement to use vtk>=9.0.

I'm actually thinking vtk>=9.1 due to the NumPy 1.24 incompatibility for <9.1. Ref #3729

[MatthewFlamm]

If we drop CI testing, we should also bump the vtk dependency version.

Is there any benefit to having this conda test now? We still use pip to install this package.

[banesullivan]

Is there any benefit to having this conda test now? We still use pip to install this package.

The conda-forge VTK recipe has a mismatch in enabled modules from the wheel on PyPI. IMO, it makes the most sense to test against the latest conda-forge VTK to ensure PyVista gracefully handles any disabled modules that we have come to expect by default from the PyPI wheel