Add version information to data locations

[nikhilwoodruff]

Fixes #104

[nikhilwoodruff]

1. Get version number as in this PR
2. Tag HF uploads with that version number
3. Set gcloud metadata with that version number

[nikhilwoodruff]

@mikesmit and @anth-volk, this is ready for re-review when you have a second. I've changed the data upload to:

1. Operate from a single function
2. For Hugging Face files, we upload the files together in one commit, and then tag that commit with the package version (we can then download the specific version with the HuggingFace API revision= kwarg)
3. For GCP files, we upload and set the metadata of each file to include the version. We can then list metadata of each blob's versions without downloading it, and download the blob with the specific version.

[nikhilwoodruff]

Just a note to double check that this script is run after versioning updates.

Confirmed that it is run after: the dataset build and upload action is triggered by changes to pyproject.toml.

[nikhilwoodruff]

Also separate the HF and GCP logic into two separate functions

[nikhilwoodruff]

@anth-volk this PR now has the changes from our in-person review just now.

[anth-volk]

Was great to go through this @nikhilwoodruff. Probably 70% of the comments boil down to: you probably want to use GH Actions's needs keyword instead of depending on changes to particular files to avoid strange edge case bugs when sequencing push actions.

[anth-volk]

One question before approval, in the comments

[anth-volk]

Let's do it