Segfault when swapping Field3D after mergeYupYdown

[johnomotani]

If mergeYupDown() has been called on a Field3D so that its yup_field and ydown_field pointers point to itself, then swapping it may cause a segfault. The problem is that the yup_field and ydown_field pointers are swapped, but continue to point to the original field. The check in the ~Field3D() destructor on whether to delete yup_field or ydown_field will then not skip the delete, and already-deleted memory will be deleted, causing a segfault.

@dschwoerer has a fix for this in 8eae21d of #956.
The issue is also fixed by #1345 which changes the implementation of mergeYupYdown().

[ZedThree]

#1345 should be going in soon -- I just need to fix the conflicts again

[ZedThree]

Fixed in #1345