Merge cassie experimental

[yangwill]

This PR updates the walking and standing controller with adjustments from the hardware experiments.

Cassie is able to reliably transition from PD setpoints to a balancing position and finally to the walking controller.

---

This change is 

[yangwill]

@yminchen, I was able to fix the end time issue when constructing the LIPM trajectory, but the fix is not entirely simple. Part of the cause is the context time being behind the robot's time and the other is because of the randomness of when each leaf system calls its discrete variable updates.

A simple solution is to always add a few milliseconds to the end times, this is slightly cleaner but still hacky.

[yangwill]

We still need a decision on how we want to deal with the start/end time when constructing trajectories. I propose enforcing the start end is at least 1ms before the end time (this way the target trajectory at the end time is guaranteed to be the desired value).