Cmfd refactor to sync with ClosedMOC/3D-MOC

[samuelshaner]

This PR will be accompanied by a PR to ClosedMOC/3D-MOC that syncs up the Cmfd, Vector, and Matrix classes and linalg functions as much as possible. Important changes in this PR include:

• Shared OpenMP locks: @geogunow suggested that OpenMP locks be shared across objects, where possible, to avoid unnecessary duplication of locks. Cmfd now has an array of locks (by Cmfd cell) that are shared with the Vector and Matrix objects used to solve the diffusion eigenvalue problem.
• splitEdgeCurrents() corrected: The procedure for splitting edge currents has been fixed to properly handle REFLECTIVE, PERIODIC, and VACUUM boundary conditions.
• Solver residual initialized to 0.: If left uninitialized, the residual output for iteration 0 is junk (at least for double precision).
• Vectorized setValues(...) and incrementValues(...) added to Vector.cpp: This implements the Cmfd shared memory optimization that @jtramm started in PR Cmfd shared memory optimization #196.
• Lattice::getLatticeSurface(...) reverted to general form: The Lattice::getLatticeSurface(...) method made some assumptions on the numbering of the surfaces and edges that is not valid for 3D surfaces, so this method was reverted to the previous general form.
• splitCorners() -> splitEdgeCurrents(): The splitCorners() routine has been changed to splitEdgeCurrents() and the nomenclature for corners has been changed to edges to reflect the language commonly used elsewhere to describe a cube:

[wbinventor]

I'll make a quick review if this after @geogunow does primary thorough review, since I've been using CMFD a great deal. I'm curious, do these.changes affect the results, e.g., the CMFD test results from #251?

[samuelshaner]

I just ran a bunch of cases from the sample-inputs, including c5g7-cmfd.py, and the test_cmfd_pwr_assembly and got the exact same results from the code in this PR and the code from PR #251.

[wbinventor]

Great! Thanks for checking @samuelshaner.

[geogunow]

Is this new spacing correct?

[samuelshaner]

All other preprocessor directives are indented to the line they address. However, the google c++ style guide says that all preprocessor directives should not be indented. This was accidental, but to be consistent with the style guide I would recommend we make a PR that removes the indent from the directives.

[geogunow]

I just read through this PR. Everything looks good aside from a couple small spacing comments. I have yet to run this on test cases. In the near future, this will hopefully be done by Travis.

[geogunow]

Per the new style, fix spacing

[geogunow]

Oops, forgot I merged this PR