Further optimize the "pow" and "powu" functions

[PaulRBerg]

A good ~350 gas can be saved in the powu function by implementing the optimizations in ABDKMath64x64. I haven't yet fully grokked how the optimizations work, but this is what I understand so far:

• Use an intermediary 64.127-bit format for the absX var
• Discriminate between the cases when x <= SCALE and x > SCALE, since in the former case overflows are not possible in the intermediary results
• When x <= UNIT, batch the checks for 2^0, 2^1, 2^2 and 2^3 in the while block, then divide by 2^4
• When x >= UNIT, progressively check if x is lower than 2^32, then 2^48, then 2^56, then 2^60, then 2^62 and finally 2^63, storing the findings in an absXShift var