Add Common ASL Raw Data Problems page to docs

[tsalo]

This would also be great to document just with some visual examples. Might be nice to have a "common ASL raw data problems" just to let users see things they might need to be aware of up front.

Originally posted by @cookpa in #516

[tsalo]

Also, folks dealing with legacy datasets often don't know what order the volumes are in (CL or LC), so a heuristic to determine that would be helpful. Try a control - label subtraction, then divide by control. For 2D data, it should be about 2%. For 3D data, should be around 3%.

[caugolm]

This would be super helpful! If you have a generally all right brain mask, the "correct" mean control-label difference within the mask should be positive and there should relatively few voxels with a negative value, while if the control/label volumes guess is flipped, the mean ought to be negative and have relatively more negative voxels