setting eb2.small_volfrac can introduce inconsistencies in EB data

[darylbond]

In my implementation, setting eb2.small_volfrac triggered a situation where the boundary normal provided by ebfactory->getBndryNormal() did not correspond with the associated face area fractions given by ebfactory->getAreaFrac(). This lead to non-conservative flux calculation when the area fraction was used to calculate the cell divergence while the boundary flux was calculated using the provided boundary normal. This situation can obviously be avoided by calculating the boundary normal from the area fractions.

[asalmgren]

Thank you for pointing this out!

PR #1451 should fix this -- if you have a chance could you check that it fixes the case you have observed?

[darylbond]

I have found another inconsistency, this time between volume fraction and isCovered(). I was able to fix this by performing a check on volfrac after every call to set_eb_data such that the covered status is consistent with whatever was just calculated. However, with further testing, it appears this approach leads to floating-point errors in EB2::coarsen_from_fine() so not such a good idea after all.

Edit: removed a bunch of code that didn't work.