Delay mutations in shuffle so shirking is correct

[TysonMN]

Workaround of hedgehogqa/fsharp-hedgehog#157 for the function shuffle.

This change also improves the runtime from quadratic to linear. The previous runtime was quadratic because of the calls to xs.Length and xs.[i] inside the loop, both of which have linear runtime.

As mentioned in the Wikipedia article, I dropped the guard i <> j because we don't need it.

Let's delay a release a bit longer. I want to see if there is anything else I can improve.

[cmeeren]

Excellent! Thank you so much. Whenever you're ready to release, just create a PR that increments the fsproj version (and optionally polish the release notes if you deem it necessary).

[TysonMN]

Whenever you're ready to release, just create a PR that increments the fsproj version (and optionally polish the release notes if you deem it necessary).

Will do.