feature: expand benchmarking

[wpbonelli]

Is your feature request related to a problem? Please describe.

FloPy currently has a very small set of benchmarks using pytest-benchmark, including

• init/load/write time for a small test model: https://github.com/modflowpy/flopy/blob/3e176d0d668675ae2372106ffe1efc829debad38/autotest/test_modflow.py#L1352
• MP7 pathline/endpoint data load time: https://github.com/modflowpy/flopy/blob/3e176d0d668675ae2372106ffe1efc829debad38/autotest/test_modpathfile.py#L264

It might be worthwhile to a) benchmark a broader set of models/utils, and b) minimize ad hoc code needed to achieve this.

Describe the solution you'd like

Maybe benchmark load/write for all test models provided by a models API as proposed in #1872, as well as any widely used pre/post-processing utils. Could also try ASV — it has been adopted by other projects like numpy, shapely, and pywatershed.

Describe alternatives you've considered

We could just stick with pytest-benchmark and a bit of scripting instead of moving to ASV.

Additional context

This would help quantify performance improvements from the ongoing effort to use pandas for file IO

[wpbonelli]

ASV seems unmaintained now, and some projects have begun to switch away from it. We could either stick with pytest-benchmark or move to something like Codspeed, with which the former would work out of the box. But in any case, probably good to set up a more complete benchmarking system in the near term so we can see the difference when we start reimplementing IO routines.