In case we find two gcodes closer than 0.01mm in Z, we estimate the layer
height from the first layers and group gcodes based on this layer height. The
logic seems fine and the code seems to work, but it's nontrivial and I'm a bit
worried that it might break things.

This is aimed at Slic3r's spiral vase mode which continuously lifts Z.

Add a specific parser for these, which can be accessed as gcoder.S/P

Modify regexps and remove ; filtering