For java code "(i >> 31) >>> 31", it can be optimized to "i >>> 31".
AArch64 has implemented this in back-end match rules, while AMD64
hasn’t.

Indeed, this pattern can be optimized in mid-end by adding some simple
transformations. Besides, "0 - (i >> 31)" could also be optimized to
"i >>> 31".

This patch adds two conversions:

        1. URShiftINode: (i >> 31) >>> 31 ==> i >>> 31

        +------+ +----------+
        | Parm | | ConI(31) |
        +------+ +----------+
           | / |
           | / |
           | / |
      +---------+ |
      | RShiftI | |
      +---------+ |
               \ |
                \ |
                 \ |
                  +----------+
                  | URShiftI |
                  +----------+

        2. SubINode: 0 - (i >> 31) ==> i >>> 31

        +------+ +----------+
        | Parm | | ConI(31) |
        +------+ +----------+
               \ |
                \ |
                 \ |
                  \ |
   +---------+ +---------+
   | ConI(0) | | RShiftI |
   +---------+ +---------+
               \ |
                \ |
                 \ |
                  +------+
                  | SubI |
                  +------+

With this patch, these two graghs above both can be optimized to below:

        +------+ +----------+
        | Parm | | ConI(31) |
        +------+ +----------+
            | /
            | /
            | /
            | /
        +----------+
        | URShiftI |
        +----------+