Use of external verified VecMap crate

[jhaye]

This does compile and should work out of the box.

But I want to make some performance testing first, before merging, since I replace every expression that uses closures.

[lhstrh]

@petervdonovan here is another use case for continuous benchmarking.

[jhaye]

Right, so there is basically no performance difference:

There are some detectable differences, but they even out overall. And they are seemingly just a result of already existing benchmark performance variation.

[oowekyala]

Let's merge this.

For future reference, I think putting the crate on crates.io is not a good idea as the crate is specifically tailored for the scheduler. There are other crates which implement similar functionality but with a higher level api, that wouldn't support some of the optimizations we do (like the Keyref thingy), like ordered-vecmap. Putting it on crates.io would increase the maintenance burden for something that is supposed to be internal.

I also am concerned about splitting the codebase in two. I thought until now that this pr was not targeting main and was something temporary for your fork @jhaye. I think maybe a good way to arrange this in the future is to move the vecmap crate into this repo using a cargo workspace.