Add support for LinearAlgebra::distributed::Vector

[peterrum]

depends on https://github.com/dealii/dealii/pull/16432

[peterrum]

Current status:

88% tests passed, 37 tests failed out of 323

Total Test time (real) = 1196.32 sec

The following tests FAILED:
	 45 - dem/find_contact_pairs.debug (Failed)
	 84 - lethe-fluid-particles/dynamic_contact_search.debug (Failed)
	 86 - lethe-fluid-particles/particle_sedimentation.debug (Failed)
	 88 - lethe-fluid-particles/restart_particle_sedimentation.debug (Failed)
	133 - lethe-particles/sliding_restart.debug (Failed)
	137 - lethe-fluid-block/apparent_viscosity_carreau_gd.mpirun=2.debug (Failed)
	141 - lethe-fluid-block/mms2d_gd.mpirun=2.debug (Failed)
	143 - lethe-fluid-block/mms2d_powerlaw_gd.debug (Failed)
	144 - lethe-fluid-block/mms3d_gd.mpirun=2.debug (Failed)
	145 - lethe-fluid-block/poiseuille3d_gd.debug (Failed)
	146 - lethe-fluid-block/poiseuille_gd.mpirun=2.debug (Failed)
	149 - lethe-fluid-block/taylorcouette_gd.mpirun=2.debug (Failed)
	151 - lethe-fluid-block/taylorcouette_gd_dimensions.debug (Failed)
	197 - lethe-fluid/interface_sharpening_blurred_circle_adaptive_mass_conservation.mpirun=2.debug (Failed)
	202 - lethe-fluid/mms-transient-conduction_gls.mpirun=2.debug (Failed)
	209 - lethe-fluid/mms2d_gls.mpirun=2.debug (Failed)
	211 - lethe-fluid/mms2d_gls_simplex_refine3.mpirun=2.debug (Failed)
	216 - lethe-fluid/mms3d_gls.mpirun=2.debug (Failed)
	220 - lethe-fluid/mms_adv_diff_diss_gls.mpirun=2.debug (Failed)
	228 - lethe-fluid/poiseuille_restart.debug (Failed)
	232 - lethe-fluid/taylor-green-vortex_gls_bdf1.mpirun=2.debug (Failed)
	234 - lethe-fluid/taylorcouette_gls.mpirun=2.debug (Failed)
	243 - lethe-fluid-sharp/check_point.debug (Failed)
	246 - lethe-fluid-sharp/coupled_moving_stokes.mpirun=2.debug (Failed)
	276 - lethe-fluid-vans/beetstra_packed_bed.debug (Failed)
	282 - lethe-fluid-vans/dallavalle_packed_bed.debug (Failed)
	283 - lethe-fluid-vans/difelice_packed_bed.debug (Failed)
	284 - lethe-fluid-vans/gidaspow_packed_bed.debug (Failed)
	285 - lethe-fluid-vans/kochhill_packed_bed.debug (Failed)
	286 - lethe-fluid-vans/pcm_packed_bed.debug (Failed)
	287 - lethe-fluid-vans/qcm_packed_bed.debug (Failed)
	288 - lethe-fluid-vans/rong_packed_bed.debug (Failed)
	289 - lethe-fluid-vans/spm_packed_bed.debug (Failed)
	293 - lethe-fluid-matrix-free/mms2d_fe1.mpirun=2.debug (Failed)
	295 - lethe-fluid-matrix-free/mms2d_fe2.mpirun=2.debug (Failed)
	297 - lethe-fluid-matrix-free/mms2d_fe3.mpirun=2.debug (Failed)
	312 - lethe-fluid-nitsche/noslip33_simplex.mpirun=2.debug (Timeout)

(note: I have not compiled with Metis)

[blaisb]

@peterrum I worked a bit on this today and it made me realize how invasive this change will be. The main issue is that the present solver really rely extensively on the fact that overloaded operator "=" when you have locally_relevant=locally_owned vectors takes care of the communication. Although I'm not the hugest fan of this, this means that the non-linear solver take care of multiple communications without necessarily knowing the vector type. The main issue will be that we will need to have the non-linear solver check if the vector being used are deal.II vectors and then have them take care of the update of the ghost. I'll check a bit more into this, but this is really more troublesome than I originally anticipated