Update API to allow users to specify case root, archive dir, etc

[mnlevy1981]

Right now the CaseClass is hard-coded to look in my scratch space for run directories and short-term archive directories. It would be nice to compare some plots to the 1 degree run, which means accessing data in Kristen's directories instead... so some sort of configuration step or additional command line argument to allow that would be useful.

[dcherian]

Would an intake catalog be the answer here? CaseClass could take a simulation_name and then look for it in the catalog.

(I've wanted to do this with the simulations I look at but haven't actually done it)

[mnlevy1981]

We're talking about creating an intake catalog for every individual CESM case we want to analyze (ideally by using https://github.com/NCAR/CESM_catalog and possibly even by building it into the CIME / CESM workflow), but so far for this project we are leaning towards hiding the intake interface unless [advanced] users want to query it directly. So we would still want some logic to figure out where to look for the catalog given a case name.

[mnlevy1981]

Some notes for future @mnlevy1981 to think about once casper and the jupyter hub are released from maintenance:

What I'm picturing now are optional arguments to CaseClass.__init__() named something like rundir_root, archive_root, and ts_root. Defaults would be '/glade/scratch/$USER', '/glade/scratch/$USER/archive', and None, and if ts_root isn't specified then the class won't even attempt to read in time series files.

A downside is that every notebook in this package would need to specify these three arguments for every CaseClass object. Another option would be to have the default values come from a .yaml or .cfg file, or to allow some such file to alter the defaults -- users could then

1. specify three optional arguments in the __init__() call
2. for arguments not specified when creating the object, look to see if they are defined in a config file
3. for arguments not specified when creating the object or defined in the config file, fallback to the defaults mentioned above

If the goal is to have intake-esm handle everything, then I'm hesitant to introduce too much complexity in what is a temporary solution... but even with the move to intake it may be useful to have these features for analyzing old runs so perhaps there's still value.

[mnlevy1981]

In #34, we ended up adding a single (required) argument: output_roots, a list that contains all potential directories such as RUNDIR, DOUT_S, or root directory for time series output (a string is acceptable as well, in that case output_roots = [output_roots] so it's treated as a list containing a single location.