Add process-specific timings

[abitofevrything]

Adds functions that use CLOCK_PROCESS_CPUTIME_ID instead of CLOCK_MONOTONIC for timings, and are therefore not affected by other processes on the system.

One thing to check before we merge: since the time is per-process, the time reported is n_threads times more than realtime since each active thread adds to the timings - should we adjust for this?

Also updated bench-all to include information about the measured parameters, better formatting and better configuration detection.

Closes #382.

[ggerganov]

@abitofevrything Can you verify that everything is good on Windows? After that I will merge it

[abitofevrything]

Tested on my Windows machine and there seems to be some issues with using clock(). Apparently its implementation is OS-specific and on Windows it returns wall clock time instead of process time.

Apparently there is a GetProcessTimes to get process time, but I'll need to look into that later.

[abitofevrything]

Well, I changed the implementation to use GetProcessTimes but the issue with the timing still persists: process time is sometimes counted as negative.

I'm pretty sure this is an underflow/overflow issue, or an issue with the wrong data type being used somewhere, but I don't have a setup to debug this. I'll see what I could do, but if anyone used to windows development has any idea what is happening here I'd be glad if you could let me know.