perf: improve code generation performance

[GabrielDertoni]

Use BufWriter and avoid formatting generated code for faster build times.

BEFORE

AFTER

This PR still doesn't address the main issue which is the compilation time of the mavlink crate itself. But for a 14 line change I don't see a reason not to do this. It makes code generation 3x faster.

PS: The measurements were taken with cargo clean followed by cargo b --test encode_decode_tests --timings on a Ryzen 5 5600X processor.

[patrickelectric]

@GabrielDertoni pls run cargo fmt.

[danieleades]

@GabrielDertoni nice one!

how much of the speedup is from buffering the writing and how much from skipping the formatting?

[GabrielDertoni]

@GabrielDertoni nice one!

how much of the speedup is from buffering the writing and how much from skipping the formatting?

@danieleades I first implemented these improvements along with a lot of other stuff I wanted to change. This is just a cherry pick. So I haven't measured in this last commit, but when I was first implementing it I did. When I measured, master took ~8s to run the build script. Then buffering took it down to ~4s and not formatting took it to ~2s.