Don't double-wrap exceptions in nurseries

[jakkdl]

Fixes #2611

I should probably remove the extraneous tests, and merge it into one of the existing test files, or at the very least rename the test file.

@Zac-HD suggested

"if we're using strict exception groups and the inner nursery raises an exception group with a single exception in it (subtree, not necessarily leaf!) unwrap it concat the tracebacks"

and I suppose that's maybe slightly different behavior from what I implemented?
I added a test case for when the child case itself raises an ExceptionGroup, should the traceback be modified in any way in that case or is it fine to just pass it through?

[codecov]

Codecov Report

Merging #2826 (e803320) into master (f85021e) will increase coverage by 0.00%.
Report is 1 commits behind head on master.
The diff coverage is 100.00%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #2826   +/-   ##
=======================================
  Coverage   99.16%   99.16%           
=======================================
  Files         115      115           
  Lines       17482    17514   +32     
  Branches     3113     3125   +12     
=======================================
+ Hits        17336    17368   +32     
  Misses        101      101           
  Partials       45       45           

Files	Coverage Δ
trio/_core/_run.py	100.00% <100.00%> (ø)
trio/_core/_tests/test_run.py	100.00% <100.00%> (ø)

[Zac-HD]

I think it's fine to pass through.

For the tests, there are currently enough separate cases that I can imagine missing that something is backwards. Could we write something parameterized, such that it obviously does each possible set of operations, and then the asserts only depend on whether we're in strict mode? (I expect this to be harder to read individually, but the confidence in consistency is what I'm after)

[Zac-HD]

This looks great to me - the tests still take a moment of thought, but suffice to convince me that it's now correct in general 🎉

The CI failures look like one flake, and one mypy issue to fix - trio/_core/_tests/test_run.py:2563: Missing type parameters for generic type "ExceptionInfo" [type-arg], I think ExceptionInfo[BaseException] would be easiest - and once that's done let's merge!

[A5rocks]

Oh just realized this should probably also contain a newsfragment?

[jakkdl]

Ah shoot, forgot the newsfragment. Opening a new PR for it