Use a kubernetes infrastructure

[severo]

As done for https://github.com/huggingface/tensorboard-launcher/ and https://github.com/huggingface/moon-landing gitaly before.

• API: see Use kubernetes #227
• workers: see Add datasets-server-worker to the Kube cluster #236
• nginx reverse proxy: see Add a nginx reverse proxy in front of the API #234
• deploy to the prod cluster: see Create the prod infrastructure on Kubernetes #249
• respond to datasets-server.huggingface.tech: see Change domain name to datasets-server.huggingface.tech #222
• add a webhook from moonlanding to the new datasets-server URL: asked on Slack
• warm the cache on all the public datasets - see the state of the cache here: https://grafana.huggingface.tech/d/SaHl2KX7z/datasets-server?orgId=1
• update moonlanding: see https://github.com/huggingface/moon-landing/pull/2957
• update the moonlanding proxy: see https://github.com/huggingface/conf/pull/166
• update autotrain: no need, autonlp-ui accesses the server through a proxy on moonlanding (https://huggingface.co/proxy-datasets-preview)
• remove the old machine
• remove the datasets-preview-backend.huggingface.tech domain
• setup monitoring: see Setup prometheus + grafana #250 and The API is unavailable in production #279
• remove old webhook: asked on Slack

[severo]

Thanks a lot @XciD for the help on the issue !!!!