This PR fixes the ilowpass off-by-one problem reported by Stuart Stevenson here: https://sourceforge.net/p/emc/mailman/message/35761624/
The intent is to make ilowpass round its internal value to the nearest integer (to be presented on its .out pin), rather than truncate. Truncation makes .out be the integer nearest to 0, as seen from the internal value.
This PR thus changes the behavior of ilowpass: it now flips from one .out count to the next sooner (when the internal value crosses the half-way mark, instead of crossing over the next integer). I think that’s generally ok, and generally more correct, but i’d like it if someone tried to think of a way this breaks existing configs.