Fixed ICE for EII with multiple defaults due to duplicate definition in nameres

[jdonszelmann]

r? @jieyouxu (since you looked at the other one)

Fixes #149982

Previously a fix was proposed by @SATVIKsynopsis which I marked as co-author on the first commit for the test they contributed. I'm closing this previous PR.

Duplicate definitions of EII defaults shouldn't be possible. I want to still panic on them, since I want to know when other bugs exist. However, in this case the duplicate was caused by something more subtle: both eiis have the same name, and as such a "duplicate definition" error is given. However, the compiler gracefully continues compiling despite that, assuming only one of the two EIIs is actually defined.

Both defaults then name resolve, and find the same single remaining EII, and both register themselves to be its default, breaking the single-default assumption.

The solution: I added a span-delayed-bug, to make sure we only panic if we hadn't previously had this duplicate definition name resolution error.

Thanks to @SATVIKsynopsis for their attempt. Adding a diagnostic here could make some sense, but nonetheless I think this is the better solution here <3
Also thanks to @yaahc for debugging help, she made me understand the name resolution of the situation so much better and is just lovely in general :3

The last commit is something I tried during debugging, which felt like a relevant test to add (one where both eiis also have the same function name)

[SATVIKsynopsis]

Thanks for the detailed explanation. that makes a lot of sense.
Glad the root cause is clearer now, and happy the tests were useful.