* Extend the benchmarks

* Optimize encoding/decoding of primitive type vectors

* Adds a benchmark for a complex type

* Optimize `u8` and `i8`

* Implement some more `size_hint` and improve array `encode_to`

* Fix the range...