Possible speed-ups for get_grouped_by_powers

[nstarman]

This function uses .append. It looks like with a simple helper function (defined externally to this function) we could use a list comprehension instead. I wonder if that would be faster — append versus function call overhead.

Originally posted by @nstarman in #16392 (comment)

itertools has a recipe for this type of dual-list creation loop. Doing two loops at the C-level might be faster than 1 loop with constant appending.