[Core] Fix norm in residual criteria and revert tests to original values

[marandra]

This PR fixes a bug in the computation of the 2-norm (missing sqrt), introduced between commits 11377ef and ba897e3

Most importantly, it reverts tolerance values used in a Structural Mechanics test to the original values, so tests are passing again. The test failed after the modification of the computation of the norm, so the test was changed. I understand that with the modification in the routine, it was expected different results, so a change in the test might be needed. This is just a remainder to be VERY CAREFUL when modifying tests, as it completely defeats their purpose.

--- a/applications/StructuralMechanicsApplication/tests/test_patch_test_truss.py
+++ b/applications/StructuralMechanicsApplication/tests/test_patch_test_truss.py
@@ -92,7 +92,7 @@ class TestTruss3D2N(KratosUnittest.TestCase):
         linear_solver = KratosMultiphysics.SkylineLUFactorizationSolver()
         builder_and_solver = KratosMultiphysics.ResidualBasedBlockBuilderAndSolver(linear_solver)
         scheme = KratosMultiphysics.ResidualBasedIncrementalUpdateStaticScheme()
-        convergence_criterion = KratosMultiphysics.ResidualCriteria(1e-8,1e-8)
+        convergence_criterion = KratosMultiphysics.ResidualCriteria(1e-16,1e-24)
         convergence_criterion.SetEchoLevel(0)

         max_iters = 1000
@@ -118,7 +118,7 @@ class TestTruss3D2N(KratosUnittest.TestCase):
         linear_solver = KratosMultiphysics.SkylineLUFactorizationSolver()
         builder_and_solver = KratosMultiphysics.ResidualBasedBlockBuilderAndSolver(linear_solver)
         scheme = KratosMultiphysics.ResidualBasedBossakDisplacementScheme(0.00)
-        convergence_criterion = KratosMultiphysics.ResidualCriteria(1e-8,1e-8)
+        convergence_criterion = KratosMultiphysics.ResidualCriteria(1e-16,1e-24)
         convergence_criterion.SetEchoLevel(0)

[philbucher]

thanks @marandra !
this PR seems fine to me, but please also wait for the response of others
also FYI @KlausBSautter since he made those tests

[jcotela]

@marandra Thanks for fixing this. I'd however ask you to wait for a while to merge, since I suspect that the trilinos version will have the same mistake. Id like to test it during the afternoon and push a fix to this branch if necessary.

[marandra]

@jcotela I suspected that it could impact MPI or other parts of the code. Feel free to merge and delete the branch when you are done.

[loumalouomega]

Ups, sorry

[jcotela]

Ok the trilinos test is still failing, but we (@philbucher and I) are not convinced that it is due to the residual criteria itself. We would merge as is and fix the test later.

[marandra]

So we merge then? We need to re-approve.

[jcotela]

Yes. Thanks for the re-approval, too.

[marandra]

Thanks!