Include training data in docker container

[ericnakagawa]

Problem: Currently the tutorial site build takes 1h40m on a g3-8x instance. Reducing this time will allow the community to interact and help us improve documentation. Target: <30m

Solution: Public training data hardly changes. Include tutorial training data in the containers used to build the tutorial site and modify the makefiles and build scripts to check for the existence of this data prior to downloading data.

Impact: 15-20min saved

Related: #315 #316 #317