Refactor __get_msb to work on a single 128bit debrujin sequence

[TomAFrench]

We're currently using a debrujin sequence for 64bit values twice to get the msb of each limb. We should replace this with a single debrujin sequence for 128 bit values

[kashbrti]

ah yes, I did this as a hack to get the refactor over the line.
will add the 128-bit version.