K2: Avoid using Nothing? as inference result in the majority of cases

Namely, do not choose `Nothing?` result type when fixing a variable
that has other constraints besides the ones that came from
the relevant type parameter's upper bounds.

See more details in KT-55691.

In K1, the case from specialCallWithMaterializeAndExpectedType.kt
was working (inferred to String?) just because the branches
were analyzed independently with `String?` expected type.

This change became necessary after the previous commit when we united
inference subsystems for if/when branches (see motivation there).

NB: For K1, the behavior is left the same, but the code
was refactored a bit.

^KT-55691 Fixed
^KT-56448 Fixed

compiler/testData/ir/irText/declarations/kt47527.fir.kt.txt

@@ -12,7 +12,7 @@ fun test_1(value: Any?): String? {
 }
 
 fun test_2(value: Any?): String? {
-  return run<Nothing?>(block = local fun <anonymous>(): Nothing? {
+  return run<String?>(block = local fun <anonymous>(): String? {
     return { // BLOCK
       val tmp1_safe_receiver: Any? = value
       when {