Additional models for Data View

Currently it is assumed that users of the Data View will write their own models. However there are going to be very common use-cases that we want to support to prevent pointless repetition of implementations. These should ideally not introduce too much complexity, and should share features where appropriate.

There is already an implementation of a multi-dimensional array data model. Beyond this it would be nice to have:

- [ ] a model that can support Pandas DataFrames and data structures which follow the same general API
- [ ] a model that supports columnar views of data (essentially cleaning-up and testing of the column example)
- [x] a model that supports tabular row views of lists of data (#661 is a first cut of this)
- [ ] a model that supports hierarchical row-based views of data

The sort of common features that we may want to develop include a full generalization of the data accessor concept in #661 for extracting values out of objects, sequences and mappings.