Major update of bBeamFormerPipeline

It now properly selects the parameters for the kernels after the beamformer, like it is done in COBALT. The new option "--beamformer" allows configuring the backend ("ccg", "wmma", or "reference"). Some other (cosmetic) aspects of the benchmark, like the performance reporting and progress bar have also been improved. Minor changes to the rest of the code were needed: 1) the virtual ~Kernel destructor was removed (to allow a std::unique_ptr to a BeamFormerCCGKernel instance), a LOFAR_RTC_VERBOSE was added to have less clutter in the output and the timing of JIT compilation (first stage of runtime compilation) was changed (it was always zero).