Testing significantly slows down with more tests

In order to reduce the impact of added tests in the same file, this PR limits the number of tests run in parallel to 4.

Without changes on the current main commit, this was the effective limit, as there were not VM tests that had more than 4 tests. But in #307 many more tests were added.

Limiting concurrent test runs to 4 should help alleviate the added time, especially for windows.