Use Ninja job pools to throttle the compilation of parsers

The following discussion from !1146 (merged) should be addressed:

• @clemenci started a discussion:

  It would be definitely better to use ninja job pools as in !1150 (merged).

@clemenci Is this trying to avoid compilation errors that happen when you run out of memory? Because DD4hep had the same problem, but they managed to find a very effective solution: https://github.com/AIDASoft/DD4hep/blob/master/CMakeLists.txt#L66

That's exactly what I wanted to do although I didn't think about the memory query trick. Thanks for the hint.

Just a note to whoever implements this. We should probably use set_property(GLOBAL PROPERTY APPEND JOB_POOLS ...) (note the APPEND) to allow the toolchain to define pools, too.