fix async113 false alarms

[jakkdl]

fixes #263

I came up with another false alarm that requires a bigger fix, so leaving that one as a tomorrow problem. It appears that 2.5 years is past my limit of fully remembering complicated details of the innards of this code base 😅

[Zac-HD]

I wonder if it'd be easier to scan the body of the nursery block searching for yields, rather than saving potential errors to emit later? Small performance hit, but much easier to implement and I'd be less worried about subtle bugs.

Separately, I'd love to test on a nursery where we yield and then start_soon a task.

[jakkdl]

I wonder if it'd be easier to scan the body of the nursery block searching for yields, rather than saving potential errors to emit later? Small performance hit, but much easier to implement and I'd be less worried about subtle bugs.

eyy, this cracked it. And with setting no_visit if there's no yield I think this is a performance gain. and it fixed another false alarm, though one that relies on pretty egregious code.
...
and it let me remove the other code that led to other missed alarms since manually visiting doesn't let the TypeTrackerVisitor interject. bishbashbosh perfect fix

Separately, I'd love to test on a nursery where we yield and then start_soon a task.

👍

[Zac-HD]

needs changelog etc, then merge 🚀