Improve clipping (problematic data) detection

[vincentvanhees]

Following error report in the google group
it seems desirably to make the clipping detection a bit more elaborate.

In the example give there the values in one axis were consistently above 2.5 for over an hour.

Proposal: Add check per axis whether any value is in the range [-2, 2] g. If not then label as clipping. This on top of the existing checks for clipping.

Here, I am trying to account for a scenario that a sensor is not well calibrated and not worn for the entire recording.