sleep jobs are not failing appropriately

[Jeff Lane ]

Looking at some test submissions, Ara caught this bug in the commands for the sleep jobs. The commands look like this:

fwts_test -l $CHECKBOX_DATA/suspend_30_cycles -s s3 --s3-device-check --s3-device-check-delay=30 --s3-sleep-delay=30 --s3-multiple=30 | tee $CHECKBOX_DATA/suspend_30_cycles_times.log

and the output looks like this:
==================== Test Results ====================
 - Cycle 0: Status: PASS Sleep Elapsed: 3.84376 Resume Elapsed: 1.64022
 - Cycle 1: Status: PASS Sleep Elapsed: 3.46496 Resume Elapsed: 1.60628
 - Cycle 2: Status: PASS Sleep Elapsed: 3.46789 Resume Elapsed: 1.69945
[SNIP]
 - Cycle 28: Status: PASS Sleep Elapsed: 3.37511 Resume Elapsed: 1.64873
 - Cycle 29: Status: PASS Sleep Elapsed: 3.37120 Resume Elapsed: 1.58720
Average time to sleep: 3.54618
Average time to resume: 1.65150
High Failures: 1
 - sleep

Note the presense of a High Failure. However, the test is marked pass.

This appears to be because checkbox is getting the exit code of the tee command rather than fwts_test. This has been seen before in other jobs, and using 'set -o pipefail' is the solution.