fix(resolve_exe): typecast exe name to string before passing to _resolve

[martclanor]

This PR aims to address #2455. The issue is related to an edge case wherein a missing executable that is being passed to resolve_exe is provided as a pathlib.Path object.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 75.8%. Comparing base (bb9824e) to head (3bdfe7f).
Report is 60 commits behind head on develop.

Additional details and impacted files

@@            Coverage Diff            @@
##           develop   #2457     +/-   ##
=========================================
+ Coverage     68.4%   75.8%   +7.4%     
=========================================
  Files          294     293      -1     
  Lines        59390   61914   +2524     
=========================================
+ Hits         40652   46972   +6320     
+ Misses       18738   14942   -3796     

Files with missing lines	Coverage Δ
flopy/mbase.py	72.6% <100.0%> (+2.8%)	⬆️

... and 254 files with indirect coverage changes

[wpbonelli]

Why does moving str() outside the internal _resolve fix it? Figured we were missing a none check somewhere.

I wonder if there is a better way to just to look up a program by name or path. Having introduced this function in the first place I regret its necessity. It is a source of frequent pain

[martclanor]

Inside _resolve, typecasting to string is already done but only for one case (out of 2 cases). I thought instead of adding str() to the other case, it might be better to just ensure that we pass a string object to _resolve. This also eliminates the need to add a test wherein we pass (to resolve_exe) a Path object of an executable that doesn't exist.

Not sure how to change it while keeping it forgiving (e.g. with the add/remove suffix logic), but I'm thinking maybe with pathlib.Path.is_file instead of shutil.which we can make it at least a bit simpler?