Refactor the inner logic of the crate to cleaner, faster and zero-copy code

[zbraniecki]

This is just the beginning, but already functional.

I got the format example to mostly work, which allows me to compare the benchmark and it gets me 57us -> 4.5us on my machine.

I'm not sure if I'll stick to the Display trait everywhere, since I want to maintain the pluggable formatters for color/no-color and potentially richer formatters of the display list in the future, but the core logic is here.

I still need to re-add a lot of features, which I'll be working on next.

If you have feedback on the proposed logic, or suggestions on how to make things lighter and faster, please, take a peak at the PR :)