Define project roadmap

Some recent observations raise the question of what the long-term project roadmap for Tilegarden should be. Some of these observations:

* Lambda cold-start times are painful, and connecting to RDS can be slow
* Fargate recently got much cheaper, so the prospect of running a containerized tile server on Fargate is more attractive
* We are evaluating a few different promising vector tile servers, some of which come with pre-built binaries or container images (see https://github.com/azavea/tegola-on-lambda-proto)
* Tilegarden needs some work to make it easier to integrate into new projects (see #121 and #135)

Two options for renewed focus might be:

* Reduce focus on Mapnik integration, instead invest in e.g. a module for provisioning CloudFront + API Gateway for a tile server
* Focus on making raster performance and deployment integration really good, so that Tilegarden might be the go-to solution for raster tile servers

Determine what the most promising aspect of Tilegarden is, and set up a project roadmap for how to make Tilegarden even better at that.

Review the deployment concerns in the issue, including Lambda, RDS, Fargate, CloudFront, API Gateway, Mapnik, and the tegola-on-lambda-proto project. Read issues #121 and #135 to understand integration gaps, then compare the two proposed directions. Done means documenting a decided focus and a concrete project roadmap.

Written by the indexing model from the issue text.