Generalize GridTools::rotate for arbitrary rotation axes.

[marcfehling]

We have the Physics::Transformation::Rotations namespace which we can use the generalize the GridTools::rotate functions. Currently, the rotation matrices in GridTools::rotate are hard-coded.

I deprecated the old interface, in which you can specify the (Cartesian) rotation axis with an integer. Further, I removed some comments at the end of grid_tools.h which fixes the indentation of #include directives. This makes this PR looks way larger than it actually is.

I adjusted all tests to use the new function with the new interface. These tests only rotated triangulation around Cartesian axes. I can provide another test with an oblique orientation axis if you want to.

Further, the Physics::Transformation::Rotations::rotation_matrix_3d takes the axis parameter first and then the angle, while the GridTools::rotate function takes the angle first and then the axis. Shall we take the chance and unify the interface?

[masterleinad]

Otherwise, looks OK to me.

[marcfehling]

Adjusted this PR to have a unified interface between namespaces GridTools and Physics. We also now use Tensor as a parameter for the axis -- this will not break any user code, since Point objects will be implicitly converted.

[masterleinad]

In file included from /Users/runner/work/dealii/dealii/source/base/symmetric_tensor.cc:20:
/Users/runner/work/dealii/dealii/include/deal.II/base/symmetric_tensor.templates.h:761:17: error: no matching function for call to 'rotation_matrix_3d'
            R = dealii::Physics::Transformations::Rotations::rotation_matrix_3d(
                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/Users/runner/work/dealii/dealii/include/deal.II/physics/transformations.h:915:38: note: candidate function [with Number = double] not viable: cannot convert initializer list argument to 'const Tensor<1, 3, double>'
Physics::Transformations::Rotations::rotation_matrix_3d(
                                     ^
In file included from /Users/runner/work/dealii/dealii/source/base/symmetric_tensor.cc:20:
/Users/runner/work/dealii/dealii/include/deal.II/base/symmetric_tensor.templates.h:765:17: error: no matching function for call to 'rotation_matrix_3d'
            R = dealii::Physics::Transformations::Rotations::rotation_matrix_3d(
                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/Users/runner/work/dealii/dealii/include/deal.II/physics/transformations.h:915:38: note: candidate function [with Number = double] not viable: cannot convert initializer list argument to 'const Tensor<1, 3, double>'
Physics::Transformations::Rotations::rotation_matrix_3d(
                                     ^
In file included from /Users/runner/work/dealii/dealii/source/base/symmetric_tensor.cc:20:
/Users/runner/work/dealii/dealii/include/deal.II/base/symmetric_tensor.templates.h:769:17: error: no matching function for call to 'rotation_matrix_3d'
            R = dealii::Physics::Transformations::Rotations::rotation_matrix_3d(
                ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/Users/runner/work/dealii/dealii/include/deal.II/physics/transformations.h:915:38: note: candidate function [with Number = double] not viable: cannot convert initializer list argument to 'const Tensor<1, 3, double>'
Physics::Transformations::Rotations::rotation_matrix_3d(
                                     ^
3 errors generated.

[marcfehling]

/rebuild

[marcfehling]

Adjusted this PR to have a unified interface between namespaces GridTools and Physics. We also now use Tensor as a parameter for the axis -- this will not break any user code, since Point objects will be implicitly converted.

Adjusting the interface for the Physics function broke the test physics/transformations-rotations_01 fixed in 70b14fb. Maybe we should leave the old Physics function then and deprecate it.

[marcfehling]

Squashed the commits. Ready for final review.

[kronbichler]

That was a funny one - only 12 digits and then with wrong rounding (the first omitted digit of pi is 9, so even in case this we would have found 12 digits sufficient it should have been ...5359) - thanks for fixing things up.

[marcfehling]

I was checking which tests are affected by the change by grepping for the function name and thought I could fix this along the way. There might be more occurrences of these magic numbers in the testsuite. I can check for these as a follow-up.

[marcfehling]

Something funny happened while adding your suggestions. PR is back on track now!