Test and improve `.model.water`

[khaeru]

This is a tracking issue for follow-up items from #88. Create 1 or more PRs and (if necessary to keep discussion organized) issues to address:

• Add tests to raise coverage to 100%. including testing for the input data (@awais307 )
• Consider methods for packaging data, e.g. omit some data from the built message-ix-models package, and provide a script to retrieve it from GitHub, Zenodo, or other location(s).
• Provide link(s) to the GMD manuscript in review or as published. (@awais307 )
• Rename the CLI from mix-models water-ix to mix-models water (after corresponding change has been made on message_data). (@adrivinca )
• Adjust the water reporting scripts to work also for sub-annual time steps (@adrivinca )
• change private_data_path to package_data_path (@adrivinca )

[khaeru]

Consider methods for packaging data, e.g. omit some data from the built message-ix-models package, and provide a script to retrieve it from GitHub, Zenodo, or other location(s).

Just for information here, I tried the following:

$ python setup.py bdist_wheel sdist

• With the added files in message_ix_models/data/water/, the size of the total (compressed) package is now 2.7 MB.
• Unpacked, those files are about 10.8 MB.

The fact that you've chosen to handle this data as CSVs rather than binary formats like MS Excel means that it compresses quite well. This is also not a lot compared to other packages (e.g. numpy is 10–20 MB compressed), so my sense is that it is not, for now, worth the effort to invent a more elaborate way of handling this data.