Add default science configurations

[SeanBryan51]

This change modifies the behaviour so that when a science configuration file is not provided by the user, benchcab will default to using internally defined science configurations.

Addresses issue #23

[ccarouge]

What is here is good. But I would add a global attribute to the netcdf output files. Maybe add the attribute "IS_DEFAULT_CONFIG" and the values could be: "DEFAULT_CONFIG_SET" or "CUSTOM_CONFIG_SET". Feel free to find better names/values.

I know the config is already listed but that's easier to see and find this way.

[SeanBryan51]

What is here is good. But I would add a global attribute to the netcdf output files. Maybe add the attribute "IS_DEFAULT_CONFIG" and the values could be: "DEFAULT_CONFIG_SET" or "CUSTOM_CONFIG_SET". Feel free to find better names/values.

I know the config is already listed but that's easier to see and find this way.

Yep we should do this for reproducibility reasons. However this made me realise, I think we should avoid having command line options that might effect the results of the test because:

• these CLI options would have to be specified in the output (as we are planning on doing now) which might clutter the netcdf global attributes over time
• in practice, it might be a pain to work out what command line arguments were used to reproduce a test (we would have to get access to the run directory and inspect their netcdf output to see how benchcab was run)
• ideally, the config file should contain all benchcab related settings so everything is reproducible solely from the config

So instead of specifying science configs as a command line argument, we add it to the config. What do you think?

[ccarouge]

I agree, the more we can put in the config.yaml, the better.

[ccarouge]

Be careful! This breaks adjust_namelist_file for now since it expects the configs without the cable level. But leave it in since that's coming with the next PR.