Regression in scv performance with axis-limited accelerations #686
Description
b0e18b3 reverted all of the DK changes to SCV and copied back the Klipper behavior of using the pre-limiting acceleration for computing JD and the post-limiting, move-specific acceleration for computing the junction limit from JD.
Prior to that, DK (pre-Kalico) had been fixed to keep everything in SCV units and not use this roundabout divide-by-one-accel, multiply-by-another. This was done to make it so that SCV performane is invariant under per-axis limiting.
Please revert these changes or re-work them on top of the more recent work to get back the lost performance.