Investigate whether we can calculate better PairPos format1 subtable splits by tracking shared subtables

[cmyr]

This is fairly low priority, since the cost of an additional subtable is low, and this particular optimization is unlikely to save us more than a split or two, but I'm curious anyway.