[lworld] RFR: 8280707: [lworld] More issues with C2's arraycopy intrinsic

[Tobias Hartmann]

The memory output of the `MemBarCPUOrder` node added to prevent writes into the source from floating below the arraycopy is not correctly wired. As a result, we hit an "dependence cycle found" assert in GCM. I also removed a comment in that code that was already removed from mainline.

In `TypeAryPtr::add_field_offset_and_offset` we set the wrong offset if the field is not found in a flat array (for example, because the inline type has no fields). This leads to an assert in the matcher due to an unexpected address type change.

Thanks,
Tobias

-------------

Commit messages:
 - 8280707: [lworld] More issues with C2's arraycopy intrinsic

Changes: https://git.openjdk.java.net/valhalla/pull/620/files
 Webrev: https://webrevs.openjdk.java.net/?repo=valhalla&pr=620&range=00
  Issue: https://bugs.openjdk.java.net/browse/JDK-8280707
  Stats: 93 lines in 3 files changed: 75 ins; 14 del; 4 mod
  Patch: https://git.openjdk.java.net/valhalla/pull/620.diff
  Fetch: git fetch https://git.openjdk.java.net/valhalla pull/620/head:pull/620

PR: https://git.openjdk.java.net/valhalla/pull/620