Remove bounds checks

[mhr3]

This branch removes some bounds checks from the for loops, which speeds up bunch of operations. I've mostly focused on the in-place methods, given the overhead of the allocations and data copies in the other methods is quite large. Also tbh I'm not really sure why does go sometimes require the slice to be a separate variable and sometimes not, but I can confirm every one of these changes is necessary to remove the bounds checks.

Here's a benchstat comparison of this branch vs master:

name                                    old time/op  new time/op  delta
BitsetOps/Equal-8                       1.08µs ± 1%  0.67µs ± 2%  -37.56%  (p=0.000 n=9+9)
BitsetOps/FlipRange-8                    700ns ± 2%   649ns ± 2%   -7.38%  (p=0.000 n=10+8)
BitsetOps/NextSet-8                      722ns ± 3%   471ns ± 0%  -34.74%  (p=0.000 n=10+9)
BitsetOps/NextClear-8                    962ns ± 3%   473ns ± 3%  -50.80%  (p=0.000 n=9+9)
BitsetOps/DifferenceCardinality-8        673ns ± 2%   656ns ± 1%   -2.51%  (p=0.000 n=10+10)
BitsetOps/InPlaceDifference-8           1.02µs ± 2%  0.51µs ± 2%  -49.76%  (p=0.000 n=10+10)
BitsetOps/InPlaceUnion-8                1.01µs ± 2%  0.51µs ± 3%  -49.38%  (p=0.000 n=10+10)
BitsetOps/InPlaceIntersection-8         1.03µs ± 2%  0.52µs ± 0%  -49.52%  (p=0.000 n=10+9)
BitsetOps/InPlaceSymmetricDifference-8  1.03µs ± 2%  0.52µs ± 0%  -49.98%  (p=0.000 n=10+10)

[lemire]

Can we fuse those in one line?

_, _ = b.set[wn-1], c.set[wn-1]

[mhr3]

I haven't checked the output of check_bce, but yeah, should be fine.

[lemire]

@mhr3 That's a fantastic PR and we will merge it. Can you have a look at my two comments? These are not requests for you to change your code, but I'd like you to react to my comments.