[WIP] Performance test with implicit function solve

[Robbybp]

The first pass at adding a performance test that solves a distillation optimal control problem with an implicit function formulation. There are several open questions here, including whether performance tests with an embedded solve even belong here.

• What are the right categories to time? I've timed (full space) instance creation, setup of the implicit function interface/model, solve, and verification of the solution.
• How to standardize tests for different test problems? I've deferred this for now, as I only have a single test problem here.
• How should dependency versions (ASL, Ipopt, CyIpopt, even NumPy/SciPy) be accounted for? I haven't included this information anywhere, but it should probably be appended to the metadata.

Note that the implicit formulation needs to use a direct solver, so there is no single file write we can compare. It would probably be possible to time individual nl file writes, but I haven't tried to break down solve times beyond the bins I mention above.

This allows a comparison like the following, where I compare two solves, one of which inserts a sleep(1e-5) call in calculate_variable_from_constraint.