FIR: Fix inference for case of non-nullable RHS of elvis

Some existing tests start failing after previous commits adding @Exact
attribute to `?:`

They have a form:
var x: String? = nullable()

if (x == null) {
   x = nullable() ?: "" // considering @Exact the whole elvis is inferred to nullable from expect type
}

x.length // should be smart cast

compiler/fir/analysis-tests/testData/resolveWithStdlib/inference/ifElvisReturn.fir.txt

@@ -14,7 +14,7 @@ FILE: ifElvisReturn.kt
                 R|<local>/b|
             }
             else ->  {
-                R|/materialize|<R|kotlin/Nothing|>() ?: ^foo Unit
+                R|/materialize|<R|B|>() ?: ^foo Unit
             }
         }
 

compiler/fir/resolve/src/org/jetbrains/kotlin/fir/resolve/transformers/body/resolve/FirControlFlowStatementsResolveTransformer.kt

@@ -243,6 +243,20 @@ class FirControlFlowStatementsResolveTransformer(transformer: FirBodyResolveTran
             }
         }
 
+        session.typeContext.run {
+            if (result.typeRef.coneTypeSafe<ConeKotlinType>()?.isNullableType() == true
+                && result.rhs.typeRef.coneTypeSafe<ConeKotlinType>()?.isNullableType() == false
+            ) {
+                // Sometimes return type for special call for elvis operator might be nullable,
+                // but result is not nullable if the right type is not nullable
+                result.replaceTypeRef(
+                    result.typeRef.withReplacedConeType(result.typeRef.coneType.makeConeTypeDefinitelyNotNullOrNotNull(session.typeContext))
+                )
+            }
+        }
+
+
+
         dataFlowAnalyzer.exitElvis(elvisExpression)
         return result
     }