Invert tmask to interpolate the correct timepoints

[dhasegan]

Discussion is here: #473

I am still debugging this so don't pull this in yet.

[dhasegan]

do I have to change tmask to ones in utils/tfilter ?

tmask=ones

to:

tmask=zeros

[dhasegan]

I tested the last version and it correctly processed my volumes (both censored and not).
I think it's safe to push this and I can chat about it if needed.

[dhasegan]

For example, my qcfc figure resulting from 125 volumes censored (using a variation of the fc-36p_scrub.dsn):

[a3sha2]

@dhasegan thank you very much,
tmask is set to 1 for any flag volumes if we change this now, this has to be changed during interpolation

[a3sha2]

that figure is correct for scrubbing outputs. It is necessary to indicate which volumes were flagged, that is why you see c two residual images in regress folder, one without scrubbed volumes and the other with scrubbed volumes

[dhasegan]

I see them, its the _residualized_dtseries.svg and _residualized_hemi-L_bold.func.svg for left hemisphere.

Looks good to me. Is there anything else that I should do for this PR?