Evaluate Using Interleave Kernel for Zip

[tustvold]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.

Currently the zip kernel makes use of MutableArrayData. This will work well when the mask contains long runs of true or false values, and when the inputs are not scalar.

Describe the solution you'd like

It is conceivable that constructing an input for the interleave kernel is always faster, but it is also possible a hybrid approach that does the following might be possible:

• Count bits in mask
• If less than some threshold set or unset use MutableArrayData
• Otherwise use interleave kernel

This will involve writing good benchmarks to assess the impact of any change

Describe alternatives you've considered

Additional context