Fix shiftedmetricinterp unit test

[ZedThree]

I'd forgotten to commit setting GlobalNzNoBoundaries in FakeMesh 🤦

As discussed in the other PR, I've also set the region for Z interpolation to be NOY

[github-actions]

clang-tidy made some suggestions

[dschwoerer]

Thanks 👍

[ZedThree]

Ah, we have found the test fails when the region manipulation in ZInterpolation isn't done -- test-laplace-petsc3d with CHECK=4 sets the boundaries to NaN, and ShiftedMetricInterp ends up hitting these.

This is coming from communicating the solution inside the Laplace solver. The solver is only applying its BCs one boundary cell deep, and we get a NaN from the second one (the outer most boundary cell). We (probably) can't just apply our own BCs here in case they conflict with the solver's, but it's possible we could just extrapolate or something to the outer boundary cells.

I think the simplest thing is to just revert those changes.