errorfill(xx, yy, dy) fills the area defined by (xx, yy ± dy).
Normally, xx, yy, and dy all are N-vectors.
errorfill(..., 'up') only plots upward; errorfill(..., 'down') only plots downward.
To specify downward and upward errors separately, dy may be shaped Nx2.
dy must be positive, even for the downward error.
errorfill(xx, yy), where yy is shaped Nx2, directly specifies the bounds as (xx, yy[:,0]) and (xx, yy[:,1]), much like matplotlib's fill_between.