Small simplification of indexmanipulations

[lkdvos]

This PR removes a (possibly micro-)optimization that bypassed the Strided.jl machinery for a small gain in performance. Here I removed that part again since it severely complicates things for the GPU implementations.
It's also safer in the sense that the current implementation was only correct since it was implicitly never called on conj'd StridedViews, so I do think that in the long term this is more maintainable anyways.

I will try and see if I can run the benchmark suite to actually gauge if there are any measurable performance implications here, but I would guess that this is a reasonable trade-off.

[lkdvos]

See also QuantumKitHub/Strided.jl#52

[Jutho]

I had never even noticed when this was introduced (or had forgotten about it). Was this the reason for the special copyto! implementation QuantumKitHub/Strided.jl#52 ?

[kshyatt]

@Jutho It's the reason for one of them, yes!

[lkdvos]

Feel free to merge without the benchmarkresults, I'm pretty sure it's not realistic for us to maintain things like this anyways.

[Jutho]

@kshyatt , what is the other one?

[kshyatt]

@Jutho It's in the logic for flip! The offending line is here

[kshyatt]

With the copyto! + Broadcasted changes in the Strided.jl PR, we can avoid scalar indexing for all the flip operations 🤠