WIP Improve performance of bitmap calculation in StgToByteCode
This optimizes computation of pointer bitmaps in StgToByteCode
We were doing a bit too much work before. This changes the order of operations, and reduces the amount of intermediate data being produced.
This is meant to be a lighter weight alternative to !8536 for improving bitmap computation performance. If this MR doesn't improve performance enough we should consider !8536 and accept the additional complexity.