C2: Loop opts are missing during OSR compilation

While benchmarking AVX related tests on x86 server machines, it was observed that the test throughput of the release build is ~35x slower than that of the fastdebug build. A sample test case named TestShortRShift.java is also attached.

Preliminarily performance analysis revealed that the fastdebug version is running AVX instructions whereas the debug version is falling back to the scalar instruction.

--------------------------------------------------------------------------
Below we list the main function that is being tested in TestShortRShift.java

    public static void mainTest(long numIterations) {
        while (numIterations > 0) {
            for (int i = 0; i < BUFFER_SIZE; i++) {
               // Code below would be run using x86 AVX instructions
                outputBuffer[i] = (short)(inputBuffer[i] >> 3);
            }
            numIterations--;
        }
    }