FSI CI Test Improvement

[eyoung55]

• Add a static flag2d mesh to read during the CI test, this avoids issues with minor version changes in Gmsh producing minor variations during mesh generations resulting in minor deviations in the solution
• Don't write the mesh again after the build() step, since it will be written during the build process as part of the write-to-read step which distributes cells equally across processes
• Add back in an assertion that the solved lift/drag must match the truth vals

closes #54

[eyoung55]

The problem wasn't exactly that there was some bug or difference between (1) building a mesh and running that simulation vs. (2) reading in a copy of the same mesh and then running the simulation. The issue was that when not reading the mesh, the process was two steps: build() and then write_mesh_files(). But this call to write_mesh_files() led to problems, since as part of building, the mesh will be written and read automatically as part of distributing the cells.

So, building looked like:

1. create geometry
2. create original mesh
3. write original mesh
4. read original mesh (redistributes cells)
5. over-write mesh (with now redistributed cells)

Such that a subsequent attempt to reproduce the same simulation read in the over-written mesh with the redistributed cells, and potential discrepancies.

[arswalid]

Changes to avoid redundant mesh writing, all tests pass. 1 conflict in test_solve.py resolved