Fragment tests are defined by a small proto file describing an API
surface with characteristics such that it is desirable to test the
generated surface for correctness or to prevent regressions.

As part of a fragment test, the generator is run on a fragment to
create a GAPIC library for the fragment. The generated unit tests
for the fragment are then executed to test the surface.

Within a session, e.g. Python 3.9, all fragment GAPICs are generated
and tested concurrently.