Resolving next turn pathfinding calculation

[real_tabasco_sauce]

We currently calculate paths immediately for only 40 units, while the remainder figure out their path in time for the next turn. What this means unfortunately is that the remaining units have to stop for a whole turn and then accelerate from 0 m/s.

 Instead, we can let units continue on their previous path so that they don't stop and have to re-accelerate. The downside here is that units will continue in the wrong direction a little if you tell them to make a 180 degree turn.

There are probably better long-term solutions, like giving large selections a really quick and dirty "guess" path before calculating the new paths.

seem like an upgrade?