suggest_bads improvements

[jdammers]

suggest_bads should be improved by adding/allowing for more parameter/options and more importantly we should also use the frequency domain to find noisy channels
Suggestion: look for large variance in the frequency range between power line peaks (e.g., 60 - 80 Hz or 70 - 90 Hz for data from the US)

[nkampel]

In general the sensitivity for PSD anomalies can be increased by decresing the sensitivity_psd parameter in:

suggest_bads(raw, sensitivity_steps=97, sensitivity_psd=95,
fraction=0.001, epoch_length=None, summary_plot=False,
show_raw=False, validation=True):

If this is an issue with american data data i would be interested in investigating the raw.fiif files.

[pravsripad]

Please check out the new examples available after updating jumeg https://github.com/jdammers/jumeg/tree/master/examples/preprocessing

The suggest_bads and interpolate_bads examples can be improved.