generate_node functions now use is_const of the new features

Classification error

See merge request tpurcell/cpp_sisso!23

Removes potential of conflicts in feat_inds as the feature set is built

The width paramter was cauing issues with that test passing

intel-bin should now work

Should be 0.5 exactly not slightly off

int overflows are leading to issues

Updates to ConvexHull1D

The new is_valid critieria changed the models it was finding

Used scores_sel and not sent_scores for combining features in SIS

Not directly equal to, but based on the difference
Reset compare function to 1e-9

Necessary to get accurate results for 1 and 1/x in some cases

prevent is_valid from calling nullptr

0.0 is min for classification

should be get_omp_max_threads not num_threads

Done to remove copy/paste from python binding functions to C++ constructors

Helps prevent copy/paste of code

This should remove some of the issues with classification finding feature combinations like
[x, 1 /x] and other convex curvature functions

Make a class structure to help remove memory leaks

The problem with it was remapping and incorrect second dimension test

From sissopp

See merge request tpurcell/cpp_sisso!22

So it actually runs the test

Changes led to a bug in the python test

It will now run

These are transferred from nl_opt_parameterization

forgot some

This speeds up building time

Changes help keep model loading in python consistent

see https://gitlab.com/sissopp_developers/sissopp
commits: up to 763e40e4deaf7a45dba3f0d8052cca053c1fa957

Build source documentation with doxyge

SEe if this will work for non virtual envs

Update docker image ci

See merge request tpurcell/cpp_sisso!19

# Conflicts:
#   .gitlab-ci.yml

seperate out unit and bin tests, remove tests from the build step