Regression test for Ewald stress tensor

[bhourahine]

Geometry is not relaxed, so tensor has finite values for all elements.

[bhourahine]

@terminationshock hopefully this should be sensitive to the your test changes in the Ewald stress, so can be a valid regression test for refactoring that routine.
While looking at this, I was reminded that the reciprocal Ewald for 1/r and 1/r2 do not use the distributeRangeInChunks() mechanism. Do you think that would also help with the timings for the Ewald use case that started this discussion of vectorising?

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (c72a10e) 70.63% compared to head (53d845b) 70.63%.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1375   +/-   ##
=======================================
  Coverage   70.63%   70.63%           
=======================================
  Files         230      230           
  Lines       43943    43943           
=======================================
  Hits        31041    31041           
  Misses      12902    12902           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[terminationshock]

Your new test is not sensitive to changes in derivStressEwaldRec. My diff is:

diff --git a/src/dftbp/dftb/coulomb.F90 b/src/dftbp/dftb/coulomb.F90
index d864717b..152292ba 100644
--- a/src/dftbp/dftb/coulomb.F90
+++ b/src/dftbp/dftb/coulomb.F90
@@ -2158,8 +2158,8 @@ contains
       sigma(:,:) = sigma + (-unity + sTmp * spread(gg, 1, 3)*spread(gg, 2, 3)) * cos(rg) * eTerm
     end do
     ! note factor of 2 as only summing over half of reciprocal space
-    recSum(:) = 2.0_dp * recSum * 4.0_dp * pi / vol
-    sigma(:,:) = 2.0_dp * sigma * 4.0_dp * pi / vol
+    recSum(:) = 1.0_dp * recSum * 4.0_dp * pi / vol
+    sigma(:,:) = 1.0_dp * sigma * 4.0_dp * pi / vol
 
   end subroutine derivStressEwaldRec

The test is still successful, though:

Test #14: dftb+_scc/GaAs_8 ....................   Passed