data collection API

[briancavalier]

Ideally, we want to keep cola.next as data-format-agnostic as possible, and currently it does a pretty good job, since we rely on a datasource's metadata descriptor to help us interact with the data. One place we haven't covered yet is an observer that needs to consume a collection of data from any datasource ... for example, a bard view.

For example, datasource.fetch() may return anything it wants ... an Array, a backbone Collection, a promise for any of the above, etc. So, an observer who wants to be able consume data from any datasource can't know what it will receive.

Here are a couple ideas:

1. Specify a minimal API that any collection must implement
2. Add an API to the metadata descriptor

In either case, the API could be something super-simple like either a forEach() method, or an iterator() method which returns an ES6-like { next: function() } iterator. Both of those allow streaming and asynchrony, so they seem flexible enough. Streaming is trickier with iterators since they are pull-based, but they do facilitate lazy evaluation a bit more easily, which could be good.