In memory compilation "benchmark"

This is not really a good benchmark of the in-memory compilers, but at least it gives a glance into where it helps.

Testing method: I used Ubuntu 16.04 WSL on my Windows 10 machine.

• for in-memory compilation measurements I ran ./compiler_backend.py, opened http://127.0.0.1:8000/ with the HelloWorld input and modified one of the constant values 7 times. I threw out the first two results (warmup) and took the average of the other 5 measurements
• for normal compilation measurements I ran ./compiler_backend.py --noInMemoryCompilation instead, everything else was the same as previously

There are a lot of problems with this of course, for example the browser only opens 6 TCP connections at once, but this does not count into the compilation time actually, and of course running 6 compilers at the same time wants to use more than 100% CPU, so they will slow down each other.

Also good to know that starting a new process in WSL is really slow. So the speed up would be much less in a native environment...

language	normal compilation	in-memory compilation	speed up
Perl	749 ms	(*) 610 ms	1.22x
Swift	4799 ms	(*) 3857 ms	1.25x
C++	7793 ms	(*) 4669 ms	1.7x
Go	4929 ms	(*) 1957 ms	2.5x
C#	6002 ms	273 ms	22x
Ruby	769 ms	33 ms	23x
Java	7342 ms	284 ms	26x
JavaScript	1535 ms	38 ms	40x
TypeScript	7843 ms	176 ms	44x
Python	844 ms	15 ms	56x
PHP	1991 ms	19 ms	105x

(*) Does not support in-memory compilation actually, but still runs faster because other compilers require less resources (eg. CPU), so it has more.