C2: Continuous deoptimizations with Reason=unstable_if and Action=none

Continuous deoptimization w/ Reason_unstable_if and Action_none happeded in an online application and significant performance drop was observed.
It seems this bug is triggered when
1. Too many deoptimizations happened in a method.
2. In the last recompilation, Reason_unstable_if Action_reinterpret uncommon traps were changed to Reason_unstable_if Action_none.
3. The uncommon traps were executed but this method never recompiled.

To reproduce:
javac DeoptimizationTest.java
java -XX:PerMethodRecompilationCutoff=4 DeoptimizationTest
PerMethodRecompilationCutoff is used to make it easier to reproduce and is not used in the above online application.