"Semi-automate" periodic gathering of HF dataset statistics

[deanwampler]

Let's start with monthly running of Joe's pipeline for gathering the stats. With such low frequency, it would be sufficient to use a calendar reminder to run it, rather than something more automated (unless that's easy to do).

[blublinsky]

Can we define the meaning of the word statistics better? Is this just license info? Something else?
The complete definition for data set cards has a lot of fields with potential info, while in reality, the majority of cards have language, license, pretty_name. Some additionally have configs, size_categories, and task_categories.
In general, data card is a yaml file embedded in readme.md. It is loaded as a dictionary and instead of predefined set of fields has a set of fields based on the yaml definition. SO just getting a list of fields in the card is an interesting statistics

[deanwampler]

We'll start with the data Joe gathered in his experiment, then decide how to expand the set.

For timing this task, it is gated by setting up the AWS environment, #104.