scheduler time in quantum underflows, leading to negative values

[derekbruening]

In testing my per-core-runqueue #6938 implementation I noticed in the logs that the current time in quantum was going negative. I tracked it down to what looks like both a voluntary switch and preempt at the same time. There is code that corrects the overshoot for the quantum due to waiting for the next instruction boundary, but its conditions do the wrong thing for the simultaneous triggers.