This sets `set -e` to ensure any errors in the test script are
propagated upwards and noticed by the Github workflow.
Prior to these changes failures in the build process of the tests would
be ignored, and the success / failure would be determined purely by the
return code of the `./problems` script.