Fix regression from #71.

[chanchiwai-ray]

Set the "Health" key to a string of "NA" when it's falsy (e.g. None, "").

Note:

d.get("key") or "NA"  # return NA if d = {"key": <falsy-value>}
d.get("key", "NA")     # return None if d = {"key": <falsy-value>}

If "Health" is None, it can cause the following error:

AttributeError: ("'NoneType' object has no attribute 'replace'", Metric(redfish_storage_controller, Storage Controller information obtained from redfish., info, , [Sample(name='redfish_storage_controller_info', labels={'system_id': 'System.Embedded.1', 'storage_id': 'AHCI.Embedded.1-1', 'controller_id': '0', 'health': None, 'state': 'Enabled'}, value=1, timestamp=None, exemplar=None)]))

[sudeephb]

Can you add a unit test covering the scenario where "Health" is not present?

[sudeephb]

Thanks, LGTM

[chanchiwai-ray]

Sorry, @sudeephb @jneo8 I added one more test case, can you check again ?

[dashmage]

Just one tiny nit on the comment, LGTM!