Load segments out of locally cached files rather than over the network

[jaredlockhart]

We recently changed our segment loading/parsing to use metric_config_parser which removes the burden of complex data definitions and validations in Experimenter, but it also introduces fetching from remote Github repos which adds potentially time consuming and/or error prone network calls to the initialization of the Python request process. Since we're already fetching and storing the segment config files from metric hub in the Experimenter repo, let's try using metric_config_parser to load those files locally to prevent network calls from taking place at process startup.

44bc841#diff-4c1722c291a9894c49eb93e4d38c83e2f9ba2384f558532d078152dd83c76d85R38

┆Issue is synchronized with this Jira Task