Add class for pulling training data and providing it tensorflow

[dylanmcreynolds]

The TagService is really just a simple API for getting tags in and out of a data, and relating them to assets in databroker.

We need a class that can call both the TagService and databroker and produce a data set ready for an ML tool like Tensorflow. This class will have a call that accepts a set of tags, a number of images to find and a folder to place them. When called, it will query the TagService, find assets with the available tags, then call databroker to get images, then write them into the output directory.

This is not a final design, but something like:

`class TagExporter():

def init(self, mongo_db, root_catalog: BlueskyRun):
...

def export_for_ml(self, tags: List[string]):
...
`