Align/fix license metadata info (#4613)

* language(s) and license(s) for consistency

* `licenses:` => `license:`

* `languages:` => `language:`

* backward-compat

* make style

* set default values for deprecated languages and licenses

* Creative Commons Attribution 2.5 South Africa License

* Do a first batch

* progress

* moar progress

* moar

* moar

* Handle other `other`

* moaaaar

* Yay!!! should be done

* licenses: use the same standard list as elsewhere

* js -> python

* style

* only allow "other" and not "other-*"

* add license_details to DatasetMetadata

* fix test

* add tsv to package_data

Co-authored-by: Quentin Lhoest <lhoest.q@gmail.com>

Commit from https://github.com/huggingface/datasets/commit/7e514c312fcc1d4b8f8e297df5549f669bfb30f8

README.md

@@ -5,9 +5,9 @@ language_creators:
 - crowdsourced
 pretty_name: Wikipedia
 paperswithcode_id: null
-licenses:
+license:
 - cc-by-sa-3.0
-- gfdl-1.3-or-later
+- gfdl
 task_categories:
 - text-generation
 - fill-mask
@@ -24,7 +24,7 @@ size_categories:
 - 10K<n<100K
 - 100K<n<1M
 - 1M<n<10M
-languages:
+language:
 - "no"
 - aa
 - ab