[SDE] Port ABOBA and OBABO dynamical integrators from StochasticDiffEq.jl#545

[ChrisRackauckas]

Migrated from PR: SciML/StochasticDiffEq.jl#545
Author: @HadrienNU

Implementation of ABOBA and OBABO Langevin integrators (GJ family), plus enhancements to the existing BAOAB integrator to support matrix-valued friction (gamma parameter). This was a substantial PR with new caches, perform_step implementations, and convergence tests.

Key changes:

• New ABOBA and OBABO algorithm structs (weak order 2)
• ABOBAConstantCache / ABOBACache, OBABOConstantCache / OBABOCache
• perform_step! implementations in src/perform_step/dynamical.jl
• BAOAB refactored to support gamma::AbstractMatrix (matrix friction) via Cholesky decomposition
• Weak convergence tests in test/weak_convergence/sde_convergence_dynamical.jl
• OBABO requires an extra noise process (alg_needs_extra_process)

Files changed: src/StochasticDiffEq.jl, src/alg_utils.jl, src/algorithms.jl, src/caches/dynamical_caches.jl, src/perform_step/dynamical.jl, src/solve.jl, test/runtests.jl, test/sde/sde_dynamical.jl, test/weak_convergence/sde_convergence_dynamical.jl

Full diff preserved at: https://github.com/SciML/StochasticDiffEq.jl/pull/545.diff

References:

• Leimkuhler B., Matthews C., "Robust and efficient configurational molecular sampling via Langevin dynamics", J. Chem. Phys. 138, 174102 (2013)
• Leimkuhler B., Matthews C., "Molecular Dynamics With Deterministic and Stochastic Numerical Methods", Springer (2015)