MergerFollowLoopers / CompressionKernels_step0 optimisation

• Run new algorithm on this 4 dimensional step (2 block sizes - 2 grid sizes)
• Best mean time found for pbpb at 50kHz: 1824.58 ms ± 10.85 ms
• When plugging the new parameters in GPUDefParametersDefaults.h and compiling: 2061.96 ms ± 29.28 ms
• There might be a mismatch bewtween RTC and normal compilation?

Problems following new commits

• Appeared invalid configurations that worked before. For some configurations of the two kernels, these error arises: 
  :0:rocdevice.cpp :3018: 1129822182886d us: Callback: Queue 0x7f263ea00000 aborting with error : HSA_STATUS_ERROR_MEMORY_APERTURE_VIOLATION: The agent attempted to access memory beyond the largest legal address. code: 0x29

  /usr/bin/rocprofv2: line 324: 3728557 Aborted (core dumped) LD_PRELOAD=$LD_PRELOAD "${@}"