GH2 correction device may throw ValueError when DAQ is not cycled after number of pulse change

Found during #169941, the GH2 correction device may throw this ValueError when the number of pulses was changed on the receiver without cycling the DAQ. This likely causes some mismatch between the advertised and true shape of data in the pipeline hashes, as it was immediately cured by cycling the DAQ.

The main problem here was the silent way of failing here, as correction devices switched into PROCESSING without any visible rate even though all inputs were properly present.

Error in 'input_handler': ValueError: cannot reshape array of size 998400 into shape (50,1280)
Traceback (most recent call last):
  File "/scratch/xctrl/sa1_cal/karabo/extern/lib/python3.8/site-packages/calng/base_correction.py", line 1271, in input_handler
    image_data.shape = expected_shape
2024-03-04T15:16:18.382 ERROR  karabo.xms.InputChannel  : Exception in 'triggerIOEvent' for instance 'SA1_XTD9_HIREX/CAL/CORRECT00_M1:dataInput': 1. Exception =====>  {
    Exception Type....:  Python Exception
    Message...........:  Error in 'input_handler': ValueError: cannot reshape array of size 998400 into shape (50,1280)
    Details...........:  Traceback (most recent call last):
  File "/scratch/xctrl/sa1_cal/karabo/extern/lib/python3.8/site-packages/calng/base_correction.py", line 1271, in input_handler
    image_data.shape = expected_shape
    File..............:  /[...]/src/karathon/Wrapper.cc
    Function..........:  void karathon::detail::treatError_already_set(const boost::python::api::object&, const char*)
    Line Number.......:  889
    Timestamp.........:  2024-Mar-04 15:16:18.381982
}

@hammerd

Hmm, that's unfortunate. Currently, number of frames is set based on the cell table size. The cell table array is not used in GH2 correction (I seem to recall it might contain garbage, we assume actual cell table usage is just 0101010101...), but still relevant for this case. In the big MR of doom, number of frames can be determined arbitrarily (typically from image data shape), so it wouldn't necessarily break of cell table and image data don't match. For other detectors, that would be a problem, still...