Version 0.21.0

It has been a long time since the last version. Next tags and release
will follow a shorter cycle.

188 upgrade datasets (#209)

* feat: 🎸 upgrade datasets to 2.1.0

* test: 💍 remove test because the dataset does not exist anymore

fix: 🐛 allow streaming=False in get_rows (#207)

it fixes #206.

Simplify cache by dropping two collections (#202)

* docs: ✏️ add backup/restore to migration instructions

* feat: 🎸 pass the max number of rows to the worker

* feat: 🎸 delete the 'rows' and 'columns' collections

instead of keeping a large collection of rows and columns, then compute
the response on every endpoint call, possibly truncating the response,
we now pre-compute the response and store it in the cache. We lose the
ability to get the original data, but we don't need it. It fixes #197.
See
#197 (comment).

BREAKING CHANGE: 🧨 the cache database structure has been modified. Run
20220408_cache_remove_dbrow_dbcolumn.py to migrate the database.

* style: 💄 fix types and style

* docs: ✏️ add parameter to avoid error in mongodump

* docs: ✏️ mark ROWS_MAX_BYTES and ROWS_MIN_NUMBER as worker vars

give reason in error if dataset/split cache is refreshing (#193)

* feat: 🎸 give reason in error if dataset/split cache is refreshi

fixes #186

* style: 💄 fix style

remove "gated datasets unlock" logic (#189)

* refactor: 💡 move gated datasets "unlock" code to models/

also: add two tests to ensure the gated datasets can be accessed

* test: 💍 adapt to new version of dummy_gated dataset

I changed
(https://huggingface.co/datasets/severo/dummy_gated/commit/99194748bed3625a941aaf785740df02ca5762c9)
severo/dummy_gated to a simpler dataset, without a python script, to
avoid having non-related errors. Also in the commit: load the HF_TOKEN
from a secret in
https://github.com/huggingface/datasets-preview-backend/settings/secrets/actions
to be able to run the unit tests

* test: 💍 fix wrong hardcoded value

* chore: 🤖 ignore safety warning on ujson package

it's a dependency of lm-dataformat, and last version still depends on a
vulnerable ujson version

* feat: 🎸 remove the "ask_access" logic for gated datasets

the new "app" tokens on moonlanding can read the gated datasets without
having to accept the conditions first, as it occurs for users.

BREAKING CHANGE: 🧨 HF_TOKEN must be an app token

Update blocked datasets (#187)

* feat: 🎸 block two more datasets

* style: 💄 sort the datasets to make it easier to maintain

feat: 🎸 upgrade to datasets 2.0.0 (#182)

fixes #181

feat: 🎸 revert double limit on the rows size (reverts #162) (#179)

feat: 🎸 truncate cell contents instead of removing rows (#178)

Add a ROWS_MIN_NUMBER environment variable, which defines how many rows
should be returned as a minimum. If the size of these rows is greater
than the ROWS_MAX_BYTES limit, then the cells themselves are truncated
(transformed to strings, then truncated to 100 bytes which is an
hardcoded limit). In that case, the new field "truncated_cells" contain
the list of cells (column names) that are truncated.

BREAKING CHANGE: 🧨 The /rows response format has changed