feat(aiohttp): add instrumentation of client requests

[md384]

Adds instrumentation of client requests for aiohttp. See https://docs.aiohttp.org/en/stable/client_advanced.html#client-tracing for aiohttp documentation on tracing hooks for client requests.

This change could arguably go in a separate AioHttpClientIntegration but will leave that for discussion.

[acnebs]

Just wanted to comment and say it would be cool to get this merged, as we are using aiohttp entirely for making client requests in our app and not at all for the webserver side of it (we use FastAPI/Starlette for that).

[github-actions]

This pull request has gone three weeks without activity. In another week, I will close it.

But! If you comment or otherwise update it, I will reset the clock, and if you label it Status: Backlog or Status: In Progress, I will leave it alone ... forever!

---

"A weed is but an unloved flower." ― Ella Wheeler Wilcox 🥀

[md384]

I would still like to see this PR merged. I think it has value for the community.

[ccastleberry]

I'd like to bump this again because I'd love to see this merged.

[antonpirker]

Hey @ccastleberry !

Thanks for bumping. We have now one more full time engineer that will spend time with the Python SDK, so we should get faster in the next couple of weeks (hopefully)

[sentrivana]

Hey @md384, thanks for this PR. I wanted to merge master into it to bring it up to the current SDK changes and run the tests, but it seems like the PR doesn't allow edits by others. Could you enable them or update the PR directly?

I'll take a look and test this locally today.

[sentrivana]

Looks great! Couple of small things need changing (most of them have to do with changes that happened after this PR was opened -- sorry for the long wait!). @md384 please let us know if you have time to look at this. Alternatively you can set the PR branch to allow edits from us and then we can take it from here.

No worries if you don't have time to do either, we can also proceed by merging this and then making the changes on top.

Thanks again!

[md384]

@sentrivana I can make these changes and will get them back to you soon

[md384]

@sentrivana some of the existing tests are failing for aiohttp for me. I think this is because we now use the aiohttp client for testing the aiohttp server integration where we try and set the sentry-trace header directly but my changes now create a span (and therefore a trace-id) for the client requests, resulting in a trace-id mismatch. I'll try and fix these up.

EDIT: I've added an extra commit to pull out the trace id that was added to the request automatically by the client instrumentation so we can compare to the trace-id on the server generated spans in the unit tests.

[sentrivana]

@md384 Thanks a lot for the super fast fixes! The code looks good to me, I'll give it a spin next week and see if we need anything additional (like docs) and then we can merge this.

[sentrivana]

Found two minor things.

[sentrivana]

Thanks for your patience @md384. I've enabled automerge so this now only needs to be rebased on master or having master merged into it to apply cleanly.