bandpass: align specification of chebychev and butterworth filters to what was reported in Figure S4 and methods

corrected HPF implementation by not downsampling the data. Note that this feature was correctly implemented in previous versions but appears to have been lost among some recent merges.

skip downsampling for bandpass filter, redefine bandpass filter as UpperFreqBound +- UpperFreqBound/3

This reverts commit e216a76a.

This reverts commit a602d86c.

This reverts commit 544cbb0b, reversing
changes made to a602d86c.

removed downsampling for high-pass filter in both scripts, thus scale 1 entropy is assessed for each frequency, with exclusive variations in the spectral content of the signal

added check for nans in input data, which can happen if coming from timelock with var trl lengths), which is converted to raw by ft_checkdata

added check for nans in input data, which can happen if coming from timelock with var trl lengths), which is converted to raw by ft_checkdata

% 181019 JQK | line 131, 249, 259, 261 287 commented
%            | line 163 verbosity changed
%            | save r parameter
%            | add option for HPF & BPF
%            | also use coarsegrainmethod for scale 1 (e.g., HPF)
% 180226 JQK | 364 ff. change encoding from sc to s to allow for nonlinear scale encoding
% 190130 JQK | filter entire time series first, then temporally segment
%            | do not require data.trialinfo, get trialsize from .trial
% 190405 JQK | integrated changes from previous scripts: adpative padding
%               length; no LPF at scale 1; changed indenting; removed
%               outdated code; corrected BP setting: at scale 1 only use HP
%               vs. LP