Drop solids early at % of thrust acceleration of Vessel

[lamont-granquist]

closes #702

retires the old burntime-left heuristic which everyone pretty much hated.

drops solids when their detached thrust acceleration (or TWR) drops below the thrust acceleration (or TWR) of the pre-detached vessel.

at 100%, detaching the boosters should result in the boosters exactly pacing the remaining core of the rocket. anything lower than that, and the boosters should fall back. this should entirely prevent detached boosters from flying past the core.

50% threshold is chosen as something fairly realistic, and in my testing it seemed to "look good". the exact optimal value is something less than 100% and is difficult to calculate outside of an optimizer simulation.

as the rest of the stack then gains thrust acceleration by no longer having to pull the boosters along, the ratio between the boosters and the core is more than 2:1 after separation, but computing it as whole-vessel-vs-separated-boosters keeps the math very simple.

the shuttle historically used something like ~37%.

and to be clear, the thrust has to drop so that the accel = thrust / (SRB dry mass+residuals) has be well under what the core is producing. so this auto-adjusts for low-thrust cores. it also means that the % of rated thrust where the boosters detach will be something more like << 5%. it takes into account that towards the end of the SRBs the mass will be very small, so the burnout TWR at rated thrust will be extraordinarily high, so the thrust needs to drop a lot. for the shuttle, using a setting of ~37% here, would result in a threshold of around ~1.8% of rated thrust where the SRBs would separate (depending on various assumptions and some historical data).