ref: Patched functions decorator for integrations

[szokeasaurusrex]

This PR introduces two new decorators in sentry_sdk.utils that we can use in our integrations to automate the checks for whether the integration is still enabled. Since these decorators use the new scopes API, adopting these decorators may simplify the change to the new Scopes API.

For example, we could replace the below sample, which uses the old Hub API:

async def _sentry_app(*args, **kwargs):
     # type: (*Any, **Any) -> Any
     hub = Hub.current
     integration = hub.get_integration(FastApiIntegration)
     if integration is None:
         return await old_app(*args, **kwargs)

    ... # rest of function body

with the code below, using these decorators:

@ensure_integration_enabled_async(FastApiIntegration, old_app)
async def _sentry_app(*args, **kwargs):
    # type: (*Any, **Any) -> Any
    ... # rest of function body

The above example is taken from #2836; the full code is available at that PR.

[antonpirker]

I really like the idea, this can make our integration's code way cleaner and easier to read!

[antonpirker]

Looks already very good. Left some comments.

[antonpirker]

Some tests are still failing. I guess the GQL you can fix easily. For the Starlette tests we need to check if the behavior changed now because of the wraps (I guess) if so, we would need to wait for a major release

[szokeasaurusrex]

Some tests are still failing. I guess the GQL you can fix easily. For the Starlette tests we need to check if the behavior changed now because of the wraps (I guess) if so, we would need to wait for a major release

If adding wraps is a breaking change, we could also remove wraps from this PR, and add it back in when we have a new major.

Update: see comment below

[szokeasaurusrex]

@antonpirker The GQL tests that are failing are themselves appear to be broken; I don't think that the fact that they are failing indicates that this is a breaking change.

The reason the tests fail is because they are doing some strange monkey patching to verify that the method we monkeypatch is still being called by the integration. I don't think we do tests like this anywhere else; looks like I added them six months ago when I was less familiar with how to best test integrations.

I think we should just delete the failing tests since they provide little value, but clearly are failing even when they should not. I will open a new PR to make this change.

[antonpirker]

Looks good! Great work.

Left some typo fix