grid vars nan values breaking aggr

At the gridding step the array math is resulting in NaN value that cause issues at the aggregation step. This may be due to a change in numpy, I'm not sure.

Evidence of this issue can be seen in a warning printed to stderr:

/data/brucef/code/yori/yori/gridtools.py:133: UserWarning: Warning: converting a masked element to nan.
  varmean[sgrididx[grid_box_idx[i]]] = np.mean( var[sortidx[grid_box_idx[i]:grid_box_idx[i+1]]] )

var is a masked array and np.mean is resulting in a masked value. However, varmean is not a masked array, so setting the masked value result of np.mean to a non-masked array numpy sets it to nan resulting in the warning.