Add a user ID column to the CSV import format

[tlater-famedly]

To track changes to users between updates, we need to be able to uniquely identify them. Since the source is authoritative, an ID from the source must be used to do this.

Currently, the CSV source re-uses the user's email address as this ID, but since this is PII, this means we cannot pseudonymise CSV entries. We should add a new column to enable this.

Things to consider:

• This will be a very breaking change, and since the tool is already used in production here and there, we might need a way to migrate.
  • This could simply be a check for existing users with the given email address as an external ID, but it would be nice to find a way to prevent us having to lug around code for this forever...
• Technically the ID column can be entirely free-form, but we may want to enforce it containing UUIDs or simple numeric IDs to prevent user error causing PII leaks.