Optimize generic is_nil

[Lastique]

Reduce the number of generated branches in the generic is_nil by using bitwise operations.

Note that a similar optimization is possible for operator==, but in that case there may be a slight reduction in performance if UUIDs are mostly unequal, and a slight speedup if they are mostly equal. If you're interested, I can add the change to operator== as well.

[pdimov]

Compilers already do this transformation by the way.

[Lastique]

Ok, you can close it then.

I just prefer to not rely on compilers, when possible. :)

[pdimov]

I'll wait for CE to have 1.86 so that I can look at the codegen (again.)

[pdimov]

Now that 1.86 is on CE (somewhat, that is, not on all compilers), I could verify that this optimization is done by approximately all GCC and Clang versions we support. E.g. https://godbolt.org/z/zPoEP6xoj.

[Lastique]

Ok, thanks.