cf aggregation support for exploiting cell-methods and cell-bounds for sophisticated aggregation on mixed variables

[bnlawrence]

Consider the following use case:

• A directory (or nested set of directories) includes data files which have variables which are available on both daily and monthly means. Currently we believe that the aggregator would say "I don't know how to join those", but actually there may be sufficient information to do this if we can exploit a combination of both cell methods, and cell bounds.

E.g. if the cell methods are the same, but the cell bounds differ, we may be able to exploit those to discriminate using a combination of both. Doing this would require a sort of "combined signature".

It would be good to establish the data model for such a combined signature first (as it would have use outside of cf-python), and then to implement support within the aggregation tools.