if our upper bound estimates are a close match, then search to find t…

…he actual

amount of fillable space on each side.

this didn't make it into the final entry (I was leery of changing it so close
to the deadline) but after testing it appears it would have improved play; it's
definitely not a sure thing because the search can catastrophically
underestimate the opponent's space.

fix several bugs in accounting for space

bug fix: when traversing chambers, include sub-chambers of the battle…

…front

fix several bugs in accounting for space

bug fix: when traversing chambers, include sub-chambers of the battle…

…front

chambers on the "battlefront" are mutually exclusive with any other c…

…hambers

you need to race to get there and you must ignore everything else,
unless everything else is bigger.

now uses "red/black chamber tree" to estimate fillable space

in both pre- and post- separation evaluations

do one extra step when estimating floodfill area

inaccurate floodfills were costing us losses

articulation vertices weren't being counted correctly

and thus neither were some of the adjacent areas.  play is much stronger
now, resulting in a 8-2-19 victory over a1k0nbot-2.16 on make testall &&
make testall2

okay, submitting this one as a1k0nbot-3.0