Add job metrics

[cmerrick]

Add singer.metrics.job_timer around calls to wait_for_export.

See https://github.com/singer-io/tap-facebook/blob/master/tap_facebook/__init__.py#L488 for an example use of job_timer.

Verify behavior by ensuring that existing tests pass and METRICS messages are emitted in the expected places.

[aiguofer]

Ultimately, wait_for_export calls Client.wait_for_export.

I see 3 ways to do it, each with their pros/cons:

1. Wrap the body of Client.wait_for_export with the job_timer context manager. This would ensure that if any new code is added that either calls this directly or calls wait_for_export, the job will be tracked. The one potential downside to this approach is if it's important to track how long update_state_with_export_info takes when Client.wait_for_export fails.
2. Wrap the entire body of wait_for_export. This would capture the time it takes to handle the ExportFailed exception. The issue would be that this won't be captured if something calls Client.wait_for_export directly.
3. Wrap each call to wait_for_export (and potentially to Client.wait_for_export). This has the same benefits and downside as 2 with the added downside that if something else were to make the call, they'd have to remember to wrap that as well.

I imagine the best way to go here is 1, what do others think?

Also, I'm not sure what would be a good job name. I'm thinking something like: "Export {} for {}".format(export_id, stream_type), but I don't know if there's a general standard used for the names.