handle database timeouts

[missinglink]

hey @orangejulius I've been noticing a bunch of my test cases failing with the message "no results returned" against the v1 code.

the error is caused when the elasticsearch client times out a query with the message "Request Timeout after 3000ms", we used to return statusCode:500 for this error but now we return statusCode:200 and features: [] with the message parroted back to the user in the geojson as res.geocoding.errors[0]

eg.

  "statusCode": 200,
  "body": {
    "geocoding": {
...
      "errors": [
        "Request Timeout after 3000ms"
      ],
...
    "type": "FeatureCollection",
    "features": []
  },

this PR is not really that nice (code wise) so that's totally open for a rewrite but at least it shows a place in the code where we may be able to handle this case and maybe retry those queries? I guess we wouldn't want to retry them indefinitely; because DOS.

thoughts?

[orangejulius]

I'm fine with this code. I'd like us to also make that case return 500 again, but there's nothing wrong with cramming this in here :) 👍