fix flagregions2.f90 testing whether cell is in region

[rjleveque]

require minimal overlap as random small number to avoid rounding error problems when region edge coincides with cell edge, causing tests to fail.

[rjleveque]

I think I made the corresponding changes to how Ruled Rectangle flagregions are handled so that determining if there is overlap of a flagregion and a grid cell will not be sensitive to roundoff error.

[mandli]

I will try and test this on the new tests and some of the examples and see if I noticed any differences.

[mandli]

I ran this through a bunch of storm surge examples that used ruled rectangles and it worked for all of them so I think we are good.

[rjleveque]

Thanks @mandli. I've also done more tests and think this is ok to merge. @mjberger and I have been investigating some strange clustering/gridding that showed up in one of the tests, but that's mostly unrelated to this fix, so I will go ahead and merge.