refactor: Simplify Parse trait by using Vec<u8> instead of AyncWriteExt

[sbernauer]

This should also improve performance for write only traffic (which is 99% of the traffic) as we can remove the overhead of async and error handling.

We (@fabi321, @bits0rcerer and @sbernauer) tested the performance during GPN22 and come to the following measurements:

Hardware	before	after
@sbernauer Laptop	16G	21G
@fabi321s Laptop	26G	28G
@bits0rcerer Laptop	78G	82G
80 core Ampere server (arm)	420G	407G

So to sum it up it's faster on all x86 machines, but somehow slower on the big arm server. I really don't know how removing async and error-handling make thing slower but I guess the x86 speedup is more important.