Use venv instead of virtualenv in test_installation

[EliahKagan]

This eliminates the test dependency on virtualenv by using the standard library venv module instead in test_installation.

This is independent of which, if either, users use for the environment that the test runner is run in. sys.executable is the Python interpreter of that environment, but once the test creates the virtual environment, it uses the virtual environment's python and pip commands for everything else. This is really the same exact approach as before, just instead of installing virtualenv as a dependency of the project, the standard library venv is used.

venv is more commonly used and more widely recommended today than virtualenv. So while users can use either and they both work well, it makes sense to test with venv instead. It is rare but possible for a Python installation not to have the venv module, if a fully working build is not installed, such as in a Docker container with insufficient packages installed on some operating systems. But this can usually be remedied by installing the appropriate packages, and because venv is a standard library module, it is reasonable to rely on its presence. (Furthermore, the virtualenv command as tested was looked up in a $PATH search and could just as well have been broken.)