Investigate BitSet optimizations

[l0rinc]

Extends #1334

Similarly to how #1480 reimplements CharSeq based on Vector to enable many constant time operations, we should consider rebasing BitSet on Vector also, to enable constant time modifications, instead of the current linear ones.

However, we should also consider the non-boxed version (see #1484) of Vector, as we're storing primitives here exclusively.

@ruslansennov, @danieldietrich?

[danieldietrich]

I think my comment on CharSeq optimization also applies here. We should stay with the current impl in favor of a better memory footprint.

From wikipedia: "A bit array is effective at exploiting bit-level parallelism in hardware to perform operations quickly." That said using a Vector as backend we loose important characteristics of our low-level implementation.

Let's stay with our current impl please.

[l0rinc]

Does the primitive-enabled Vector change the picture (i.e. same memory footprint, with constant add, instead of linear one)?

[danieldietrich]

BitSet will probably not part of Vavr 1. We want to focus on the most frequently used collections. Maybe we will release some kind of add-on collections package later.