#600 - Specialized support function computations

[schillic]

See #600.

julia> @btime ρ($[1., 0.], $(BallInf(zeros(2), 1.)))
  7.835 ns (0 allocations: 0 bytes)  # this branch
  87.346 ns (2 allocations: 192 bytes)  # master

[mforets]

Cool stuff 🎉

For BallInf, a similar "trick" can be applied to the support vector: instead of using radius_hyperrectangle(H) which allocates (because it has a fill), radius_hyperrectangle(H, i) is better since it can directly use B.radius.

[schillic]

EDIT: Oh, now I see that you wrote support vector 😊