Improve index construction performance with bitmaps

[brandonwillard]

What behavior of the library made you think about the improvement?

One of the internal improvements we have at .txt involves a significantly more efficient representation of the vocabulary subsets stored in the index. By replacing the use of standard sets with something like https://pypi.python.org/pypi/roaringbitmap, outlines could have the same improvements.

How would you like it to behave?

No response