Running only the unit-tests of select crates on some platforms is
problematic. We are unlikely to update this list of crates as we
introduce new ones. It is a better default to run the tests of all
crates on all platforms and selectively exclude the ones that can't run
because they are unsupported.