Allow BoxedLcpConstraintSolver to have secondary LCP solver

[jslee02]

Problem

Dantzig LCP solver sometimes returns inaccurate solution or even NAN values when the LCP doesn't have a solution. More importantly, BoxedLCPConstraintSolver uses the solution without any sanity checks, which easily leads to unstable simulation. This is reported in #892.

Solution

Let BoxedLCPConstraintSolver to use more robust LCP solver such as PGS when it fails to solve with the primary LCP solver. In addition to that, as a final resort, it discards NAN value solution when the secondary LCP solver failed as well.

Testing

Added the regression test provided by @pchorak. Thanks!

---

Before creating a pull request

• Document new methods and classes
• Format new code files using clang-format

Before merging a pull request

• Set version target by selecting a milestone on the right side
• Summarize this change in CHANGELOG.md
• Add unit test(s) for this change