Fix intermittent failures in zpool_destroy_004_pos

The test case has multiple zfs send/recv scenarios.
The scenario that was commonly failing was as follows;
1.) zfs send poolA | zfs recv poolB
2.) wait
3.) destroy -f poolA
4.) declare an error if the exit status of the send/recv pipeline
was success.

Step 4 was attempting to assert that the destruction of poolA interrupted
the zfs send. The problem is that the zfs send could complete before the zfs
destroy -f poolA was issued. If this occured then the receive half of the
pipeline would succeed and the exit code of the pipeline would be zero.

It is not possible to manage this race from a shell script.

Changed test to allow the scenario described above to not declare an error
if the zfs receive succeeds.

Submitted by: Dave Baukus
Reviewed by: asomers
Sponsored by: Spectra Logic Corp