Sema: Fix regression with 'Failure' type witness inference hack

We have a special hack to infer the 'Failure' type witness in
'AsyncIteratorProtocol' by considering the type of 'next()'
witness.

In eaf06ea I added a check to
fix some assertions that could happen if 'next()' was witnessed
by a declaration in a protocol extension or superclass, which
has a different generic signature.

However my check was too narrow, because it also prohibited
this form of inference when 'next()' was in a different
extension of the same nominal type.

Allow this again.

Fixes #79367.
Fixes rdar://problem/145341658.

Author: slavapestov

lib/Sema/AssociatedTypeInference.cpp

@@ -2496,7 +2496,11 @@ AssociatedTypeInference::computeFailureTypeWitness(
     if (!isAsyncIteratorProtocolNext(witness.first))
       continue;
 
-    if (!witness.second || witness.second->getDeclContext() != dc)
+    // Different extensions of the same nominal are OK, but if the witness is in
+    // a protocol extension or a superclass or something, give up.
+    if (!witness.second ||
+        witness.second->getDeclContext()->getSelfNominalTypeDecl()
+            != dc->getSelfNominalTypeDecl())
       continue;
 
     if (auto witnessFunc = dyn_cast<AbstractFunctionDecl>(witness.second)) {

test/AssociatedTypeInference/issue-79367.swift

@@ -0,0 +1,15 @@
+// RUN: %target-typecheck-verify-swift
+
+// https://github.com/swiftlang/swift/issues/79367
+
+// 'Failure' type witness inference should still take place when
+// the 'next()' witness is in a different extension than the
+// conformance.
+
+struct AsyncIteratorImpl<Element>: AsyncIteratorProtocol {}
+
+extension AsyncIteratorImpl {
+  func next() async throws -> Element? {
+    fatalError()
+  }
+}