Allow site developers to customize the ATLAS ingestion pipeline

[jacobwegner]

As a follow on to #7, prepare_db assumes that all possible annotations are ingested, and that we're always starting from 0.

We likely want to provide something at the site level so that developers can customize the pipeline (via a hookset or something like the python social auth pipelines functionality).

This issue and #7 may also be useful in further decoupling the "ingestion" process from the deployed ATLAS. E.g., we run scaife-viewer/scaife-search-indexer on a really powerful VM so we can parallelize ingestion of ElasticSearch documents.

There may be a use case where developers / curators work together to "bake" an ATLAS data set (whether into a Docker image or as a tarball, etc) that could be packaged and deployed with a backend.

We may also want to decouple "extraction" from "ingestion"; e.g., we extract data sets as standoff annotation.