Add test to ensure consistent constraints in parallel

[kronbichler]

This is the test needed for #14905. I added it as a separate PR as it can currently run (but would fail to produce a reasonable output). The idea of the test is to set up constraints for a DoF owned at process with index 0, but defined only from information on processes with rank 1 and 2. The problem to be represented is that neither of process 1 and 2 have the full information on the constraints, and only their combined form represents the valid state.

[bangerth]

This is ok to merge. Out of curiosity, what is the situation you were considering that triggered this whole thing?

[bangerth]

(Ok to merge = once #14905 is merged and this is rebased.)

[peterrum]

Is the output already generated with #14905?

[kronbichler]

Yes, this uses the code of #14905.

[peterrum]

Are you sure that the output is up to date? Where is Working on rank coming from and why do not all processes output Output 0, Output 1, and Output 2?

[peterrum]

The output should be:

DEAL:0::Output 0
DEAL:0::Output 1
    0 1:  0.500000
    0 3:  0.500000
    0 2:  0.500000
    2 3:  1.50000
DEAL:0::Output 2
    0 1:  0.500000
    0 3:  1.25000
    2 3:  1.50000

DEAL:1::Output 0
    0 1:  0.500000
    0 2:  0.500000
    2 3:  1.50000
DEAL:1::Output 1
    0 1:  0.500000
    0 3:  0.500000
    0 2:  0.500000
    2 3:  1.50000
DEAL:1::Output 2
    0 1:  0.500000
    0 3:  1.25000
    2 3:  1.50000


DEAL:2::Output 0
    0 1:  0.500000
    0 3:  0.500000
DEAL:2::Output 1
    0 1:  0.500000
    0 3:  0.500000
    0 2:  0.500000
    2 3:  1.50000
DEAL:2::Output 2
    0 1:  0.500000
    0 3:  1.25000
    2 3:  1.50000

[peterrum]

@kronbichler Could you rename the file to tests/lac/constraints_make_consistent_in_parallel_02.cc so that the AC::make_consistent_in_parallel() are at the same place!?

[kronbichler]

Could you rename the file to tests/lac/constraints_make_consistent_in_parallel_02.cc so that the AC::make_consistent_in_parallel() are at the same place!?

I will adapt the output once #14905 is merged. Regarding the name, I used the mpi/constraints_consistent_01.cc as a template, so we have a conflict. I can move that file to the lac subfolder, too, as it feels that all the three tests should be in the same place.