[wpimath] Add Sleipnir Java bindings

[calcmogul]

The wrapper includes reverse mode autodiff, the Problem DSL, and the optimal control problem API. I wrote it by directly translating the upstream API and tests to Java (i.e., copy-paste-modify).

I replaced the ArmFeedforward and Ellipse2d JNIs with implementations using the Sleipnir Java bindings. Switching dev binary JNIs to release by default sped up wpimath test runs from several minutes to 7 seconds.

[SamCarlberg]

This would be better as a static, named factory like VariableMatrix.columnVector(int rows)

[calcmogul]

The intent was to be consistent with the EJML matrix API, in the same way the C++ API mirrors Eigen.

[calcmogul]

One big problem this PR still needs solved is how to track the Variables created by intermediate expressions, and how to ensure their native handles are freed in a timely manner so the JVM doesn't OOM. try-with-resources isn't applicable to those, and Cleaners aren't guaranteed to run since the JVM doesn't know how much space native objects take up.

The only thing I could think of was a factory object, whether that's Problem for free variables in addition to decision variables, or a separate factory class for the autodiff use case. Each Variable would need to keep track of its source factory so new intermediate Variables can be allocated from it. Try-with-resources on the factory would clean things up.