Add trajectory accumulator to convergence routines, for the simulator to obtain real paths.

The top-level convergence routines in ics_cobraCoach encode a fair amount of path and motormap selection logic and (move, expose, calculate) sequencing. For the "move" part they eventually call some form of (calculatePath, movePath). The simulator wants the actual trajectory, evaluated along the final motor map.

A 'con discussion led to agreement that refactoring would be tricky, and so to optionally pass some trajectory accumulator object in to the top-level convergence routines. That would be passed down, eventually reaching the calculus.calculateSteps() family of functions where the evaluations could be added to the trajectory. Actually, for the benefit of the simulator, the path segment would be evaluated at some fine spacing and not just at the end. Existence of the trajectory object would also short-circuit some routines, to avoid moves and exposures, etc.

We should merge the current u/chihyi/cobracoach to master before doing this. chihyi? chyan?