missing decomposition constraint for finite-area

raised during discussion with @Mattijs - the outside perimeter of a finiteArea can be split on a processor boundary. This means loss of the BC and replacement with a bad processor definition.

@Sergio

This is the corner case (quite literally) discussed with Mattijs:

Assuming we have specified a finite-area film on the bottom wall, with its edge inflow condition on the lower left. This edge condition corresponds to the transition between wall and inlet polyPatches.

Set up and searching works fine, unless the domain decomposition coincides exactly with the finite-area edge. From the proc0 perspective, the edge is not the connection between wall and inlet, but wall and processor1.

• Effectively loses the boundary condition entirely.
• addressing errors. Proc0 has a finite area and an edge. No other procs have that edge, so can't really expect things to go well for the solution!

Need to sync information about the faces attached to the proc boundary edges globally.

Note, use this issue as general tracking for making finiteArea mesh in parallel (EP1518).

It exposes a variety of issues with finiteArea (ie, using it in ways that were not originally planned)

mentioned in commit 097b7425

mentioned in commit be848d74

mentioned in commit e6e6378e

mentioned in commit ba64e89c

mentioned in commit 055daf30

mentioned in commit 533fdd9c

mentioned in commit bd116ae1

mentioned in merge request !457 (merged)

mentioned in commit 3fc4ad6f

mentioned in commit 4214a955

mentioned in commit 0a241800

mentioned in commit 8659d34b

mentioned in commit b4c29f99

mentioned in commit 2e3e0002

mentioned in commit b5588335

mentioned in commit 357d3550

mentioned in commit aad6c1ba

mentioned in commit 829ef8b8