[cmake] Optimize compilation performances

[Me-k-01]

Fix issue #235

This PR aims to add precompiled headers to the Fitting and the SpatialPartitioning module, to accelerate the build time on the CI.

This module uses precompiled header parsing and explicit template instantiation to accelerate the total build time of multiple compile target (like the CI tests, or the code coverage test)

Changes made by this PR

• Add PCH to Fitting containing all of the commonly used fitting types (a new lib and target must be added in the cmake)
• Use Explicit instantiation to instantiate commonly used Fitting class in the CI test  
• Add PCH to SpatialPartitioning for the commonly used KdTree and KnnGraph
• Use Explicit instantiation to instantiate commonly used SpatialPartitioning class in the CI test  
• Add PCH to Common containing the Commonly used Point types
• Use Explicit instantiation to instantiate commonly used Point class in the CI test  
• Add PCH to std library (to speed up CI test parsing of <array>, <random>...)

Troubleshooting notes

When testing the compile target linkage to the precompiled header with :

g++ -O2 -I . -I Ponca/precompiled/Fitting -I ./external/eigen -c tests/src/fit_cov.cpp -o fit_cov.o
ar rcs fit_cov.a fit_cov.o
nm fit_cov.a | c++filt | grep Basket

                 U vtable for Ponca::Basket<Ponca::PointPositionNormal<double, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<double, 3>, Ponca::SmoothWeightKernel<double> >, Ponca::MeanPosition, Ponca::CovarianceFitBase>
0000000000000000 V vtable for Ponca::Basket<Ponca::PointPositionNormal<double, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<double, 3>, Ponca::SmoothWeightKernel<double> >, Ponca::PrimitiveBase, Ponca::MeanPosition, CovarianceFitTwoPassesBase>
                 U vtable for Ponca::Basket<Ponca::PointPositionNormal<double, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<double, 3>, Ponca::ConstantWeightKernel<double> >, Ponca::MeanPosition, Ponca::CovarianceFitBase>
0000000000000000 V vtable for Ponca::Basket<Ponca::PointPositionNormal<double, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<double, 3>, Ponca::ConstantWeightKernel<double> >, Ponca::PrimitiveBase, Ponca::MeanPosition, CovarianceFitTwoPassesBase>
                 U vtable for Ponca::Basket<Ponca::PointPositionNormal<long double, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<long double, 3>, Ponca::SmoothWeightKernel<long double> >, Ponca::MeanPosition, Ponca::CovarianceFitBase>
0000000000000000 V vtable for Ponca::Basket<Ponca::PointPositionNormal<long double, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<long double, 3>, Ponca::SmoothWeightKernel<long double> >, Ponca::PrimitiveBase, Ponca::MeanPosition, CovarianceFitTwoPassesBase>
                 U vtable for Ponca::Basket<Ponca::PointPositionNormal<long double, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<long double, 3>, Ponca::ConstantWeightKernel<long double> >, Ponca::MeanPosition, Ponca::CovarianceFitBase>
0000000000000000 V vtable for Ponca::Basket<Ponca::PointPositionNormal<long double, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<long double, 3>, Ponca::ConstantWeightKernel<long double> >, Ponca::PrimitiveBase, Ponca::MeanPosition, CovarianceFitTwoPassesBase>
                 U vtable for Ponca::Basket<Ponca::PointPositionNormal<float, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<float, 3>, Ponca::SmoothWeightKernel<float> >, Ponca::MeanPosition, Ponca::CovarianceFitBase>
0000000000000000 V vtable for Ponca::Basket<Ponca::PointPositionNormal<float, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<float, 3>, Ponca::SmoothWeightKernel<float> >, Ponca::PrimitiveBase, Ponca::MeanPosition, CovarianceFitTwoPassesBase>
                 U vtable for Ponca::Basket<Ponca::PointPositionNormal<float, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<float, 3>, Ponca::ConstantWeightKernel<float> >, Ponca::MeanPosition, Ponca::CovarianceFitBase>
0000000000000000 V vtable for Ponca::Basket<Ponca::PointPositionNormal<float, 3>, Ponca::DistWeightFunc<Ponca::PointPositionNormal<float, 3>, Ponca::ConstantWeightKernel<float> >, Ponca::PrimitiveBase, Ponca::MeanPosition, CovarianceFitTwoPassesBase>

We do find the U (undefined) symbol in the output, which validates that the compiler is using the extern templated class instead of recompiling it for the target.
However, the computation time increases compared to not using explicit instantiation, this might be because the inlined function are too costly to be preinstantiated by the compiler (further testing in progress).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.03%. Comparing base (4cffe9d) to head (3d99e5a).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #241      +/-   ##
==========================================
- Coverage   94.07%   94.03%   -0.05%     
==========================================
  Files          83       83              
  Lines        3323     3300      -23     
  Branches      267      265       -2     
==========================================
- Hits         3126     3103      -23     
  Misses        196      196              
  Partials        1        1              

Files with missing lines	Coverage Δ
tests/common/kdtree_utils.h	82.81% <ø> (ø)
tests/common/testUtils.h	100.00% <ø> (ø)
tests/src/barycenter.cpp	70.00% <ø> (ø)
tests/src/basket.cpp	98.97% <ø> (ø)
tests/src/binding.cpp	98.38% <ø> (ø)
tests/src/change_basis.cpp	97.36% <ø> (ø)
tests/src/cnc.cpp	95.29% <ø> (ø)
tests/src/curvature_plane.cpp	96.42% <ø> (ø)
tests/src/fit_cov.cpp	95.31% <ø> (ø)
tests/src/fit_line.cpp	96.55% <ø> (ø)
... and 15 more

... and 4 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[nmellado]

• Add precompiled Eigen target file

This is not relevant, as Eigen use templates to optimize mathematical operations you have in your code, so you cannot pre-compile it.