Deploy tile server as standalone service in RGD instance

[aashish24]

No description provided.

[banesullivan]

On the Heroku deployment, we are seeing occasional request timeouts for tile serving. This is due to some images not being preformatted as COGs or requiring a little long to fetch from remote storage. The requests in the main instance of the app will time out after 30s. Often, we'll see the the initial fetch/tiling takes longer than 30s.

To mitigate this issue, we might benefit from setting up a new tile serving microservice.

Because of Heroku's limits on requests, this may need to be done through an EC2 instance.

This means we will have multiple heterogeneous HTTP servers: further complicating deployment. Honestly, I have no idea how we would route specific requests (e.g. all api/geoprocess/* endpoints) through to the separate server but it is possible?

Once we set this up, then we could also do this for the sumbsampling endpoints to step away for sending a POST request that runs a job then having the user hit a different GET endpoint to access the produced data.

[banesullivan]

This is coming up again after ResonantGeoData/RGD-ScrumBoard#34

Requests are timing out after viewing one of the GRIB datasets