Parallel example bug fixes

[dylan-copeland]

• [x] Fixed a bug in the Options class, by correctly setting max_basis_dimension. Before this fix, max_basis_dimension was set to the FOM dimension. In parallel, it was possible for the FOM dimension to be smaller than the number of snapshots, resulting in a failure in scalapack.
• [x] Fixed multiple examples so that they run in parallel, including dg_advection_global_rom, linear_elasticity_global_rom, poisson_global_rom, poisson_local_rom_greedy.
• [x] Examples nonlinear_elasticity_global_rom, mixed_nonlinear_diffusion seem to have been fixed in parallel by PR 145, aside from issues resulting from the Options class bug.
• [x] Added relative error calculation in poisson_global_rom.
• [x] Laghos/rom regression tests pass (except for three randomized tests that fail even for the master branch).

Note that dg_advection_local_rom_matrix_interp has serial usage of frequencies.txt. For now, it only works in serial (this PR does not parallelize it).

[chldkdtn]

@dylan-copeland, visit visualization for prom of dg_advection_global_rom does not properly update the time history solutions for parallel case. Would you mind fixing it?

Also, the name of visit file for prom is the same as the name of visit file for FOM simulations.

[chldkdtn]

@dylan-copeland are there any DMD examples with broken parallel functionality?

[dylan-copeland]

@chldkdtn The -visit issues are fixed now. The DMD examples all worked already in parallel, and their sample runs already have mpirun -n 8. There are no issues in parallel for DMD, because those examples simply sample snapshots from a code that orginally worked in parallel.

[chldkdtn]

@siuwuncheung and @pranav-vempati let's review this PR.

[chldkdtn]

@siuwuncheung and @pranav-vempati let's review this PR.

[dylan-copeland]

@siuwuncheung and @pranav-vempati