ForLoopsLowering: Use last-exclusive for-loops for optimized until

progressions instead of decrementing "last".

#KT-41352 Fixed

compiler/testData/codegen/bytecodeText/forLoop/forInReversed/forInReversedReversedUntil.kt

@@ -19,26 +19,25 @@ fun box(): String {
     return "OK"
 }
 
-// JVM non-IR uses while.
-// JVM IR uses if + do-while. The surrounding "if" gets optimized in this test (constant condition), except for Long.
+// JVM non-IR optimizes out all reversed() calls. However, this has a bug (KT-42533).
+// JVM IR optimizes out the 2nd reversed() call.
 
-// 0 reversed
 // 0 iterator
 // 0 getStart
 // 0 getEnd
+
+// JVM_TEMPLATES
+// 0 reversed
 // 0 getFirst
 // 0 getLast
 // 0 getStep
-
-// JVM_TEMPLATES
 // 2 IF_ICMPGE
 // 1 IFGE
 // 3 IF
 // 1 LCMP
 
 // JVM_IR_TEMPLATES
-// 2 IF_ICMPLE
-// 1 IFGT
-// 1 IFLE
-// 4 IF
-// 2 LCMP
+// 3 reversed
+// 3 getFirst
+// 3 getLast
+// 3 getStep

compiler/testData/codegen/bytecodeText/forLoop/forInReversed/forInReversedUntil.kt

@@ -19,26 +19,25 @@ fun box(): String {
     return "OK"
 }
 
-// JVM non-IR uses while.
-// JVM IR uses if + do-while. The surrounding "if" gets optimized in this test (constant condition), except for Long.
+// JVM non-IR optimizes out all reversed() calls. However, this has a bug (KT-42533).
+// JVM IR optimizes out the 2nd reversed() call.
 
-// 0 reversed
 // 0 iterator
 // 0 getStart
 // 0 getEnd
+
+// JVM_TEMPLATES
+// 0 reversed
 // 0 getFirst
 // 0 getLast
 // 0 getStep
-
-// JVM_TEMPLATES
 // 2 IF_ICMPLT
 // 1 IFLT
 // 3 IF
 // 1 LCMP
 
 // JVM_IR_TEMPLATES
-// 2 IF_ICMPLE
-// 1 IFGT
-// 1 IFLE
-// 4 IF
-// 2 LCMP
+// 3 reversed
+// 3 getFirst
+// 3 getLast
+// 3 getStep