This automatically-generated pull request adds the commits necessary to
make the 4.1.0 release.

Once the PR is merged, mlpack-bot will tag the release as HEAD~1 (so
that it doesn't include the new HISTORY block) and publish it.

Or, well, hopefully that will happen someday.

When you merge this PR, be sure to merge it using a rebase.

Changelog

• Adapt HardTanH layer (Updated hard_tanh and added to layer_types #3454).
• Adapt Softmin layer for new neural network API (Adapt softmin layer #3437).
• Adapt PReLU layer for new neural network API (Fix PReLU #3420).
• Add CF decomposition methods: QUIC_SVDPolicy and BlockKrylovSVDPolicy (Add Block Krylov SVD method in CF #3413, Add Quic_SVD method in CF #3404).
• Update outdated code in tutorials (Update outdated codes in cf.md and datasetmapper.md #3398, Fix codes and improve doc #3401).
• Bugfix for non-square convolution kernels (Fix convolution for non-square kernels #3376).
• Fix a few missing includes in <mlpack.hpp> (Add preprocess.hpp include and add missing includes to mlpack.hpp #3374).
• Fix DBSCAN handling of non-core points (Fix DBSCAN handling of non-core points #3346).
• Avoid deprecation warnings in Armadillo 11.4.4+ (avoid deprecation warnings in Armadillo 11.4.4+ #3405).
• Issue runtime error when serialization of neural networks is attempted but MLPACK_ENABLE_ANN_SERIALIZATION is not defined (Issue error when serializing ANN models without MLPACK_ENABLE_ANN_SERIALIZATION #3451).