Public demo deployment of datalab

[ml-evs]

We should probably make a public deployment that we can direct people to for trying things out. At least initially, this could be used in talks and as demos to people who are interested, but could grow into some free service that people could actually use, co-located with other deployments (can even envisage a peer mechanism that distributes the free datalab database and storage across a series of contributing deployments).

Some things to raise sub-issues about, that would need to be solved first:

• Magic link sign-in #469: Easier authentication than ORCID/GitHub OAuth. Potential plan: magic link style login tied to contact email. This can hook into our current mechanism for saving/rotating/tombstoning API keys per user (also good for harvesting emails -- not that we want them for anything beyond sending datalab-related alerts in the future). This will also need the bit of UI I was working on in another PR (cannot link) to expose "Login via GitHub/email etc..."
• GDPR/data rights notice #470
  • probably should be a configurable field per deployment -- can add it to the metadata
• Add a file upload size limit #471
  • (for the public deployment at least)
• Limit to number of cells/samples per user #472
  • Can be pretty arbitrarily high but shouldn't allow batch mode to create 1e50 samples
• ChatBlock limits: for the public deployment, either disable or heavily reduce chatblock functionality, in particular, probably use a separate API key from our main deployment with an even lower budget cap.

Ideal timeline, by next week...

Can use this issue to track anything else might come up with making a relatively unrestricted public demo (cc @jdbocarsly)