We are slower than linear_map::LinearMap and heapless::LinearMap

[yegor256]

According to our benchmark, we are slower than linear_map::LinearMap when map size is bigger than 5, but we shouldn't be, since we use the same algorithm as they do. They use Vec, while we manage our own array. Let's investigate and fix.

[yegor256]

@alexkazik we can close this one, I believe?

[alexkazik]

The benchmark here has improved, yes. Though when running it locally (arm64) it's not that a big improvement (and very different to yours to begin with).