Add benchmark tests for common operations

[toumorokoshi]

A common concern in instrumentation is the performance hit on applications. To help us quantify and measure how well we're doing there, we should add some benchmark tests that run as part of the CI.

There is a pytest module that could help here: https://pypi.org/project/pytest-benchmark/

[rajibmitra]

I would like to work on this. @toumorokoshi

[toumorokoshi]

@rajibmitra sounds great! I think just getting the pattern established would be a good first task.
I just took a quick look and there's not a lot of machinery for what I'm thinking, but my thought just getting a timing marker on a common task: extracting / injecting trace information. Probably exercising a complex case of tracecontext extraction/injection:

https://github.com/open-telemetry/opentelemetry-python/blob/master/opentelemetry-api/src/opentelemetry/context/propagation/tracecontexthttptextformat.py

We don't have any targets right now, so I think just having the timing run during CI (by adding to tox) would be great.

[mauriciovasquezbernal]

@rmitr are you still interested in working in this one?

[mauriciovasquezbernal]

As a part of #514 I did some research and I think pytest-benchmark is a good framework to use. It integrates nicely with pytest and provides some useful features to make the implementation of performance tests easier.

[codeboten]

Follow up to ensure its part of GA

[ffe4]

Here also a link to a related PR open-telemetry/opentelemetry-specification#748

[codeboten]

Closing in favour of #1767