Adding a comment so I can learn a bit about pull requests

Update GLMdenoisedata.m

getting up to date

If the 4D data is provided in an XYZxTime format, with no zeroes, that is the data was masked and then all the zeros removed the dramatically reduce memory usage:

data1_vec = reshape(data1, [],t_dim);
mask_ind = mask(:,:,:,1) ~= 0;

data1_vec_reduced = data1_vec(mask_ind,:);

and the user provides the original mask image in opt.reconmask, this will now reconstruct the 3d images from the reduced vector at the very end, thus generating all of the pretty plots for quite data evaluation. 

Also, this will now plot the timecourses of the PCs beneath the PC maps - though it is not as elegant as Kendrick's method.

Added a bit about the opt.reconmask to the help section and changed the subplot dimensions so that the PC weight map is larger relative to the PC regressor plot.

If a logical mask is provided in the opt.reconmask field, the code will
now vecotorize the data prior to the heavy lifting, and then, using the
earlier commit, put things back together. This is cheaper than buying
more ram.

Made the PC figures slightly prettier, and tested the auto vectorizer on
my own data. Appears to work, but further validation is needed.

This reverts commit 687fccc.

***WIll break other options that use masks (HRFfitmask, etc)***
Also, uploaded example data by mistake. removed.

If the user provides other mask in opt (hrffitmask, brainexclude,
pcR2cutoffmask) then the code should now vectorize those as well - but I
haven't tested it extensively.

Git isn't hard...its just that I don't know what I'm doing. Now, this
will change add in a few more loops to vectorize masks if they are
provided. As I mentioned previously, not extensively tested.

Auto vec merge