function made to compute the maximum gradient of the histogram to detect
strange variation of samples distribution. The maximum gradient is normalized
by an extremum gradient where all samples are zeros, thus the gradient
is 1 / delta(x), we normalized the trace before processing to ensure the min, max
of the trace is -1/+1. delta_x should not be changed and is equal to 0.01 (200 steps
from -1 to 1). It is mainly designed to reject clipped trace, where
samples amplitudes have common values

Input:
trace: trace to be processed
first_sample, last_samples: samples windowing the trace to be processed
flag_plot: 0 or 1, of you want to plot

Output:
max_gradient: maximum gradient of the histogram

Example:
max_gradient=maxgrad_histo(trace,1,numel(trace),1)