AArch64: generate_compare_long_string_same_encoding and LARGE_LOOP_PREFETCH alignment

After JDK-8268231, there is a following code shape:

    if (SoftwarePrefetchHintDistance >= 0) {
      __ bind(LARGE_LOOP_PREFETCH);
        __ prfm(Address(str1, SoftwarePrefetchHintDistance));
        __ prfm(Address(str2, SoftwarePrefetchHintDistance));

        __ align(OptoLoopAlignment);
        for (int i = 0; i < 4; i++) {
          __ ldp(tmp1, tmp1h, Address(str1, i * 16));
          __ ldp(tmp2, tmp2h, Address(str2, i * 16));
          __ cmp(tmp1, tmp2);
          __ ccmp(tmp1h, tmp2h, 0, Assembler::EQ);
          __ br(Assembler::NE, DIFF);
        }
        __ sub(cnt2, cnt2, isLL ? 64 : 32);
        __ add(str1, str1, 64);
        __ add(str2, str2, 64);
        __ subs(rscratch2, cnt2, largeLoopExitCondition);
        __ br(Assembler::GE, LARGE_LOOP_PREFETCH);
        __ cbz(cnt2, LENGTH_DIFF); // no more chars left?
    }

I believe the intent is to make sure that LARGE_LOOP_PREFETCH jump target is aligned. In other words, should probably be like this:

      __ align(OptoLoopAlignment);
      __ bind(LARGE_LOOP_PREFETCH);
        __ prfm(Address(str1, SoftwarePrefetchHintDistance));
        __ prfm(Address(str2, SoftwarePrefetchHintDistance));

This is the form that every other use of align(OptoLoopAlignment) takes. Current shape probably has some minor performance penalties.

Tentatively assigning to Wang Huang who did the original JDK-8268231.