Experiment with performance of arrayvec, tinyvec and smallvec against std::Vec

[kirillbobyrev]

ArrayVec is pretty fast and gives a solid performance gain (though unclear how much before the engine is more or less complete).

However, it requires setting a maximum threshold for the capacity, which means that it doesn't optimize for the average case. smallvec and tinyvec crates can help with that.

The proposed experiment is to run 4 versions of the engine:

• Fully replace all current ArrayVec instances with Rust's std::Vec (base)
• Use ArrayVec as much as possible (closest to the current setup, likely to be optimal)
• Replace all ArrayVec usages with smallvec and also replace std::Vec whenever possible
• Same as above, but with tinyvec

It should be unblocked whenever engine reaches Proof of Concept milestone.