Update icetea tests to include stderr output of failed command

[cmonr]

Description

When icetea fails pytest testing, it's error message is not useful.
An example: https://travis-ci.org/ARMmbed/mbed-os/jobs/480215039

Having debugged this for most of the day, it was found that the icetea command that was generating JSON to feed into the test was failing due to dependency resolution caused by a recent pyOCD release (0.15).

Issue request type

[ ] Question
[ ] Enhancement
[x] Bug

[cmonr]

@ARMmbed/mbed-os-test-team FYI

Can provide more details if requested since resolution of the outstanding issue (pyOCD dependency) is still in progress: #9389

[ciarmcom]

Internal Jira reference: https://jira.arm.com/browse/MBOCUSTRIA-550

[OPpuolitaival]

@cmonr it is a bit complex way how icetea is used but the message is there:

E Exception: Error with python /home/travis/build/ARMmbed/mbed-os/tools/run_icetea.py --compile-list --mcu K64F --toolchain GCC_ARM --tcdir /home/travis/build/ARMmbed/mbed-os/tools/test/run_icetea/TEST_DIR_HW,
E return code: 1,
E error message: ,
E output:Traceback (most recent call last):
E File "/home/travis/build/ARMmbed/mbed-os/tools/run_icetea.py", line 284, in
E icetea_json_output = icetea_tests(options.target, options.tcdir, options.verbose)
E File "/home/travis/build/ARMmbed/mbed-os/tools/run_icetea.py", line 177, in icetea_tests
E list_json = json.loads(stdout)
E File "/opt/python/2.7.14/lib/python2.7/json/init.py", line 339, in loads
E return _default_decoder.decode(s)
E File "/opt/python/2.7.14/lib/python2.7/json/decoder.py", line 364, in decode
E obj, end = self.raw_decode(s, idx=_w(s, 0).end())
E File "/opt/python/2.7.14/lib/python2.7/json/decoder.py", line 382, in raw_decode
E raise ValueError("No JSON object could be decoded")
E ValueError: No JSON object could be decoded
E [ERROR] No JSON object could be decoded
tools/test/run_icetea/run_icetea_test.py:38: Exception

[cmonr]

@OPpuolitaival Hmm, I guess so.

I suppose I was expecting stderr/stdout to also be printed, instead of needing to manually run the subcommand to find the root cause of the problem. Closing since I don't feel strongly about a change being needed.