…kbook.

Couple of issues:

1. Cosima Cookbook seems to allow searching for coordinates: eg (`st_edges_ocean = cc.querying.getvar(experiment, 'st_edges_ocean', session,
                                    start_time=start_time, end_time=end_time, n=1)
st_edges_array = st_edges_ocean.expand_dims({'yu_ocean': u.yu_ocean, 'xu_ocean': u.xu_ocean}, axis=[1, 2])`
This doesn't seem to be easily achievable with intake (Cell 11 pre-update)? I can't currently figure out how to do this.

2. Updating from `cosima_cookbook` => `intake` seems to massively increase compute time for `topographic
_slop_magnitude` calculation:
`UserWarning: Sending large graph of size 47.19 MiB.` => `Sending large graph of size 123.9 MiB.`.

The source of issue 2 is currently unclear to me & is dramatically increasing computation time.

suggestions.

  - Where possible, all direct file reads have been removed - still
    requires a bit of a hack to get the grid file (Cell [10]) &
    `st_edges` (Cell [13]).
  - Varnames lowercased.
  - Typos fixed.
  - Direct indexing of catalog.
  - Redundant xarray calls fixed.

Notes:
  - It appears the catalog has several data frequencies available for
    this experiment, unlike `cosima_cookbook`.
    - I made the choice to disambiguate to 3 month freq. to minimise
    	computation size (lowest frequency available).
    - Migrating from cosima_cookbook => intake substantially increased
      dask_graph layers - was necessary to call dask.optimize() to
      flatten these out. All chunking, etc. is the same as it was -
      currently not clear to me where the discrepancy in compute graph
      size originates.
  - Currently doesn't appear that intake supports any good way to search
    for coordinate variables? Leads to aforementioned hacks.

- Extracted depth_slice & data freq to constants & changed default data
  frequency to 1 month (`1mon`)
- Added a bunch of comments & explanation surrounding chunking.
- Added some comments surrounding operations & what they're doing