[NI] Don't consider Nothing-constraint as proper to complete call

Follow-up of 9b3e17f0. There we decided to complete call if a type
 variable from a return type has proper lower constraints, now we refine
 this rule wrt `Nothing`-like constraints to avoid inferring type variables
 to Nothing, which is quite useless

 #KT-30370 Fixed

compiler/testData/codegen/box/inference/lastExpressionOfLambdaWithNothingConstraint.kt

@@ -0,0 +1,30 @@
+// !LANGUAGE: +NewInference
+// WITH_RUNTIME
+
+inline fun <T> foo(f: () -> T): String {
+    return (f() as? Inv<T>)?.result() ?: "Bad"
+}
+
+class Inv<T> {
+    fun result(): String = "OK"
+}
+
+fun <K> create(): Inv<K> = Inv()
+
+fun test(b: Boolean): String {
+    return foo {
+        if (!b) {
+            return@foo create<String>()
+        }
+
+        if (b) {
+            create()
+        } else {
+            null
+        }
+    }
+}
+
+fun box(): String {
+    return test(true)
+}