This executor is very useful when you are running tests on a slow
(single-threaded) QEMU emulator via ssh (as is the case in our CHERI fork)
since you can't use the -j flag of lit to speed things up.
In the past I have used this as a basic regression test when adding new
compiler features to check that no new LLVM assertions are triggered.
I originally added this executor for the old test format to the CHERI fork
of libc++. This change is a simple port of the executor to the new format.
I have since added wrapper scripts that start multiple QEMU instances and
distribute the testcases across these instances, but this executor is still
useful and in our CI jobs we use this as an first-stage executor for
libunwind/libcxx to avoid the long-running testing via SSH when there are
obvious bugs in the compiler.