re-think our CI testing for PIConGPU

[psychocoderHPC]

We need to re-think how we perform compile tests in PIConGPU.

PR #5042 is reducing the number of tests but this is fare away from bringing us into an accepted test time. With #4723 we added PIConGPU unit tests. More code should be tested there and stuff tested there should be removed from our examples/test/benchmarks combination in cmakeFlags

[psychocoderHPC]

A possible approche for particle solvers. Instead of testing if we can compile different solver with cmakeFlags we could simply add more species with the different solvers activated.