[enh] measure response time with more details.

[dalf]

First step to fix #162
The shown statistics are the same, except in memory there is a lot more information.
The second steps would be to re-implements UI. See https://github.com/dalf/searx/tree/metrology

Notes :

• there is no tests
• the stats are per unix process, so if there more than one as uwsgi does, it won't be accurate ( Inaccurate statistics #199 ). The current implementation has already this issue.
• the idea of implementation.py is to add a redis / sql / whatever backend later. This would fix the previous note, except we have to take into account the speed. One idea : recording a measure only records an action to do later. A dedicate thread reads all these actions and executes them. Note : two unix processes equals two dedicated threads to record measures.

[pointhi]

To share data between uwsgi threads we could propably using SharedArea?

[dalf]

I'm not sure about the score stats.

Also I saw a small issue with request time measure per engine : it's include the call to the request fonction (since it's based on params['started'] ).

Perhaps the per engine measure should also include the time out / error requests, after all if one engine has 20% of timeout each time the user has too wait for that.
But here there is an issue : how to get the params['started'] in the function wrapper (the one with the try / catch ).

After bring back the graph is not too difficult, nearly done actually except I would like to double check the consistency.

@pointhi : I don't know this projects why not. The only : measure response time mustn't be slow.

[dalf]

In hurry, I mess up : I pushed the metrology2 branch to asciimoo/searx instead of dalf/searx and added my settings.yml...

By the way, there are the graphs in /stats2
Warning : all measures have to be checked, double checked, there is always tricky parts...

[asciimoo]

no problem, force push allowed on non-master branches =)
..and I'll check it, thanks

[return42]

@dalf what do you think about this PR? .. if it has a value, it might be better tagged as RFC .. otherwise close?

[dalf]

I've started to rewrite this PR because many things have changed since the time I wrote that PR.
I close this one and will open another PR,.

Also, I would like to keep the https://github.com/searx/searx/tree/metrology2 branch until the new one is pushed.