FE: don't run overload-by-lambda-return-type with unsuccessful candidates

Related to KT-52927

compiler/fir/resolve/src/org/jetbrains/kotlin/fir/FirOverloadByLambdaReturnTypeResolver.kt

@@ -95,6 +95,7 @@ class FirOverloadByLambdaReturnTypeResolver(
         call: T,
         candidates: Set<Candidate>,
     ): Set<Candidate>? where T : FirResolvable, T : FirStatement {
+        if (candidates.any { !it.isSuccessful }) return candidates
         val lambdas = candidates.flatMap { candidate ->
             candidate.postponedAtoms
                 .filter { it is ResolvedLambdaAtom && !it.analyzed }

compiler/resolution/src/org/jetbrains/kotlin/resolve/calls/KotlinCallResolver.kt

@@ -244,8 +244,9 @@ class KotlinCallResolver(
                 }
             }
             if (callComponents.languageVersionSettings.supportsFeature(LanguageFeature.OverloadResolutionByLambdaReturnType) &&
-                candidates.all { resolutionCallbacks.inferenceSession.shouldRunCompletion(it) } &&
-                kotlinCall.callKind != KotlinCallKind.CALLABLE_REFERENCE
+                kotlinCall.callKind != KotlinCallKind.CALLABLE_REFERENCE &&
+                candidates.all { it.isSuccessful } &&
+                candidates.all { resolutionCallbacks.inferenceSession.shouldRunCompletion(it) }
             ) {
                 val candidatesWithAnnotation = candidates.filter {
                     it.resolvedCall.candidateDescriptor.annotations.hasAnnotation(OVERLOAD_RESOLUTION_BY_LAMBDA_ANNOTATION_FQ_NAME)