This patch adds initial CMake caches that set up LLDB's test suite (or at least the API test suite)
to run against the different simulators on macOS.
The goal is to give people a straightforward way to run the tests on their local machine
against the simulators. Once we have a buildbot that runs the simulator tests and we start
telling people to fix regressions, the idea is that we can just point everyone to these CMake
cache files to locally reproduce.
Also in case we actually have people bisect issues the idea is that these CMake caches are always
resulting in a functioning test suite targeting the respective simulator.
There is one really ugly part in this patch which is the manual addition of -j1 to the LIT test args.
The reason here is that simultaneously launching several simulator binaries on the same machine
seems to lead to random failures (usually the simulator complaining that the simulated device
is already booted). Until this issue is fixed the only way to get a reasonable test result is to have
the -j1 hack to prevent simultaneous launches. Of course this won't work as soon as someone runs
several test suites at the same time (or manually overwrites the LIT args), but we're working here
with what we have :)
I should point out that the buildbots will probably not use these caches as we plan to
reuse a single LLDB build for all simulator types.