Handle climo files with separate chronology

[rod-glover]

Fixes #57

The fix itself was simple. Modifying the tests to replace completely non standards-compliant test files cgcm.nc and cgcm-tmin.nc with standards compliant ones took most of the work.

[rod-glover]

I grouped related joins and filters, but this may not seem to others, as it does to me, a clearer way to organize the query.

[corviday]

     Searches the database for all files from a given model and
     emission scenario and returns a data value for that particular
     timestep (e.g. January [1], summer [14] or annual [17]) in each
     file.

Maybe I'm misunderstanding how this query works now, but I think this docstring in data.py could use an update. You'd now pass 0 for January (with timescale=monthly), 2 for summer (with timescale=seasonal) and 0 for annual (with timescale=yearly)? Am I understanding that correctly?

[corviday]

Seems to produce plausible data and accept the specified inputs. Looks good to me.

[jameshiebert]

Do we do any exception testing here? E.g. what happens if you give time_idx=1 for timescale='yearly', etc.?

[corviday]

Here's what happens at present.

The frontend data validation functions check to see (line 93) if the backend returns a string object instead of a JSON object to catch that sort of thing.

There may be a better way to deal with this, but that's what happens now.

[jameshiebert]

Yikes, yeah we definitely want to avoid the backend returning 500s!

[jameshiebert]

LGTM!