Understand impact on RFs sensitivity of time peak position when configuration of the camera changes

LST camera configuration includes a parameter that can shift the pixels trigger within the window, resulting in a shift in the "peak" time distribution between runs.

Experiments with the MC show that the RFs are sensible to the time shift and performances drop, especially at low energy in case of mismatch between MC and real data time shift. Yet, the RF doesn't use the intercept parameter, which should be the only parameter whose value is linked to the "absolute" position of the peak times.

Task list:

• Review RFs features
• Assert impact of random seed on RF performances
• If random seed effects are excluded:
  • review other possible effects
  • evaluate effect of a "time shift" in-variance processing (eg. re-scaling by image weighted average time)