RBF: add linear polynomial term in RBF interpolation

[edoardolombardi]

The pull introduces a regularization linear multivariate polynomial to the RBF basis during RBF interpolation.

The polynomial is based on the btipit class ReconstructionPolynomial and specialized only for a first degree polynomial.

Configurations of RBF nodes leading to undetermined linear systems, because identical rows/columns provided by polynomial basis terms, are automatically managed by reduction of the variables of the multivariate polynomial.

A custom class derived by RBFKernel has now to override new pure virtual methods aimed to the initialization of the polynomial, the identification of the independent variables and the evaluation of the polynomial basis.

A new integration test is added: test_RBF_00004.

[roccoarpa]

Just optional, if you can, please improve the test 00004 considering variable support radii instead of a homogeneous one.

[andrea-iob]

You have mixed tons of cosmetic changes with the functional changes associated with the pull request. It's hard for me to understand what are the real changes. I'll leave the review to @roccoarpa and @kgkara.

[andrea-iob]

Please squash together all the commits before merging.

[edoardolombardi]

You have mixed tons of cosmetic changes with the functional changes associated with the pull request. It's hard for me to understand what are the real changes. I'll leave the review to @roccoarpa and @kgkara.

Ah, yes, sorry for that. It was clang-format enabled for modified code. If you want I can try to isolate format changes with implementations.

[edoardolombardi]

The new forced pushed branch should be cleaned of formatting modifications.
Let me know if you'd like to add your review, @andrea-iob, thanks.