Sort YAML output in a more semantically useful order

[agbuckley]

At least when I've been running it recently, the YAML output seems to be sorting dict keys alphabetically, which is not very helpful when checking and iterating the output.

For example, additional_resources, by definition the non-core aspects, gets listed first in the submission.yaml blocks, ahead of e.g. name, description, location, data_file and keywords (one of many orders that would feel more intuitive to me; right now the name that defines the block comes last!).

Within individual tables' YAML files, dependent_variables comes first: it would be more natural at top-level to list the x-axis independent_variables first, then the y values, imho. Within the variables blocks, the header and qualifiers come first (before values) which seems good, but within each value the errors are listed before the actual (and more compact) value.

I've not checked the code doing the writing, but in Python 3.6+, dictionaries should iterate/persist in insertion order. So if the insertion code in the writer were to operate in more semantic orders and the key sorting were disabled, I hope the output would "automatically" become more user-friendly. Thanks!

[GraemeWatt]

Good point, this is something that always annoyed me, but I didn't see how to avoid it. It's always been a feature of PyYAML that yaml.dump writes keys in alphabetical order. Investigating now, there's a drop-in replacement called oyaml written to address this problem, but PyYAML 5.1 (released in 2019) introduced an option sort_keys (default True) to switch it off. There's an open issue (since 2020) to discuss whether sort_keys=False should be the default in some future PyYAML release, but for now we could just add sort_keys=False as an option when calling yaml.dump from hepdata_lib. This could also be done in other parts of the HEPData code (such as the converter) that use yaml.dump.