Fix a pathological case for 2d EB

[WeiqunZhang]

What could happen is a cell might be cut a tiny bit at a corner such that 3
faces with an area fraction of one and one face with an area fraction of
almost one, and the volume fraction is one. In that case, the boundary area
and centroid have been set to wrong values based on an incorrect assumption.

Checklist

The proposed changes:

• fix a bug or incorrect behavior in AMReX
• add new capabilities to AMReX
• changes answers in the test suite to more than roundoff level
• are likely to significantly affect the results of downstream AMReX users
• include documentation in the code and/or rst files, if appropriate

[ax3l]

For the 1.e-15 and 1.e-30 values in this file, you might want to consider using <limits>: std::numeric_limits<amrex::Real>::epsilon?