[ML] Anomalies table shows wrong actual and typical values for time_of_day/week detectors

[peteharverson]

For detectors which use time_of_day or time_of_week functions, depending on the time of the anomaly, and the time zone configured for the client, the anomalies table can report the wrong times for the actual and typical values.

In addition, the times for the actual and typical values are formatted in a 12 hour, rather than 24 hour clock.

e.g. for an anomaly with a record timestamp of Feb 10, 2016 @ 22:00:00.000, the table shows an actual value of 11:02, where it should should be 22:02 (using 24 hour clock) and also actual time displayed at 11:02pm is 1 hour out from what it should read for 10:02pm.

Fixes elastic/ml-cpp#413

[elasticmachine]

Pinging @elastic/ml-ui