IPC Implementation / Performance Improvements

[nithinp7]

Implements ideas from the paper "Penetration-free Projective Dynamics on the GPU", Lei Lan et al.

• Adaptive timestep strategy to guarantee penetration-free simulation.
• Modified collision constraint - restitution-based formulation that emulates barrier forces in a position-based context.
• Implement specialized cubic polynomial root-solver described in paper
• Implement A-Jacobi solver described in paper to replace (relatively) slow direct PD solver (Cholesky factorization + backsubstitution)

Implements crucial performance improvements

• Parallelizes sections of the solver using OpenMP - hides latency cost of refactorizing system matrix to include collision constraints.
• CUDA implementation of tetrahedral strain / volume constraint projection.
• CUDA implementation of CCD, especially the modified Newton method used for the cubic polynomial root-finding?
• More...?