Move side vertex average normal implementation from edge2 to edge

[GiudGiud]

as MOOSE has a test using edge3
refs #4257

[moosebuild]

Job Coverage, step Generate coverage on 63dba73 wanted to post the following:

Coverage

	739e36	#4292 63dba7
	Total	Total	+/-	New
Rate	65.00%	65.01%	+0.00%	100.00%
Hits	76957	76963	+6	1
Misses	41434	41429	-5	0

Diff coverage report

Full coverage report

This comment will be updated on new commits.

[roystgnr]

The "Test mac" failure is a PETSc environment issue and unrelated. The "Distributed make check sweep (odd)" failure is that weird intermittent MetaPhysicL error, unrelated.

I'm glad they got my suspicions up, though, because this code is wrong in its new context! Consider the Edge3 from point(0)==(0,0) to point(1)==(1,0) through point(2)==(0.5,0.25): normalizing (and possibly sign flipping) point(1)-point(0) would have been correct (e.g. returning (1,0) at s==1) for an Edge2 without point(2), but on this Edge3 it will still return (1,0) on side 1, even though with that curvature the new normal is (1/sqrt(2),-1/sqrt(2)) there.

[GiudGiud]

yeah I m still in finite volume land here. We have one test using Edge3, but mind you it s a straight one.

Ok I ll find something that works better, at least for a lagrange mapping

[moosebuild]

Job Min gcc on 9f31363 : invalidated by @GiudGiud

fluke?

[GiudGiud]

This should be better now! Thanks for catching that.

[GiudGiud]

forgot to register the test..

[GiudGiud]

should be good now.

[GiudGiud]

I did this geometrically with the mid point.
I m going to check this using the FE now

[GiudGiud]

ok no luck. I m seeing that the computation with FE disagrees with this simple approximation.
I should have expected it though, the FE quadratic functions dont have 0 derivatives at the edge nodes

[GiudGiud]

ok this works!

[jwpeterson]

Just a couple minor comments but if @roystgnr wants to merge as-is that's fine.

[jwpeterson]

(General comment which I can address separately from this PR.)

So 27 is the ElemType enumeration value here. @roystgnr maybe we should update the xda comment that we typically write to make it clear that the first number is an ElemType rather than the vague "type" and the node ids aren't optional? At least I think that's what we were trying to imply with the parentheses?

[GiudGiud]

btw I generated this manually. I could not figure out the script (generated an empty file)

[moosebuild]

Job Min gcc on 8a1623c : invalidated by @jwpeterson

Invalidate now that the PETSc on Min gcc has been updated

[moosebuild]

Job Test mac on 8a1623c : invalidated by @jwpeterson