aligator-next: new (parallel, block-sparse) Riccati solvers, CHOLMOD integration, split first- and second- order derivatives

[ManifoldFR]

This PR:

• Splits first- and second- order derivatives (Resolves #144, @fabinsch )
• Adds tracy as a profiling tool (Resolves #120)
• Upgrades the minimum version of C++ to 17 (Resolves #119, @ManifoldFR )
• Adds a new sublibrary called gar to factorize LQ subproblems for solvers (Resolves #118, @ManifoldFR )
• Adds a parallel, block-sparse factorization for the implicit/proximal Riccati algorithm (Resolves #156, @ManifoldFR )
• Integrates the CHOLMOD solver from the SuiteSparse library into gar (Resolves #148)
• Upgrades the minimum required version of eigenpy to 3.4
• Add a C++ example and benchmark of Talos walking by @edantec
• Major rewrite of SolverProxDDP : no longer use QFunction or ValueFunction structs, delegate Riccati recursion to gar solvers
• Rewrite the Logger class (rename from BaseLogger)
• Merge struct ODEData into ContinousDynamicsData, rename continous-base.hpp header to continuous-dynamics-abstract

I would like to thank the contributors to this PR: @fabinsch, @edantec, @jorisv

Resolves #155

[jcarpent]

Great achievement @ManifoldFR! Bravo!

[ManifoldFR]

Adding this comment here for future reference (it'll be added to the docs)

At time of writing, in the current master branch of Eigen (3.4.90), the global variable which tracks whether allocations are allowed is thread-safe: https://gitlab.com/libeigen/eigen/-/blob/master/Eigen/src/Core/util/Memory.h#L96

But on released versions of Eigen it is NOT, hence any check for mallocs when multithreading is enabled will break as on/off switches for the malloc status will create race conditions.

@edantec

[ManifoldFR]

@jorisv @edantec @fabinsch you guys think we can merge this bad boy ?