Handle empty user and participant names

[renefs]

• Handles empty names on avatars for users and participants.
• Adds a name_not_empty validation in the participant schema.
• Adds a migration to set to "?" the existing empty participant names

Fixes #502

[ThiefMaster]

DB constraint is maybe a bit too much for it, how about just rejecting it in the schema?

def not_empty(value):
    if not value:
        raise ValidationError('This field cannot be empty.')

[github-actions]

This PR contains database changes. Before merging it, make sure to apply the migration in production:

BEGIN;

-- Running upgrade 30b5fe7b5949 -> cab5d47c1152

UPDATE participants
        SET name = '?'
        WHERE name= '';;

UPDATE alembic_version SET version_num='cab5d47c1152' WHERE alembic_version.version_num = '30b5fe7b5949';

COMMIT;

When reviewing the PR, make sure that the changes will not break the previously deployed
version, i.e. any new column needs to have a server_default or be nullable.

[renefs]

I wonder if it's worth to add anything to the downgrade code like:

def downgrade():
    op.execute(
        """
        UPDATE participants
        SET name = NULL
        WHERE name = '?';
        """
    )

Since this is fixing a "bug"... probably not?

[ThiefMaster]

I wonder if it's worth to add anything to the downgrade code like:

No, it's a fix, so it makes sense to leave the downgrade step empty.

[ThiefMaster]

For the broken comment about the alembic revision here in this PR, can you try removing these 3 lines?

https://github.com/indico/newdle/blob/master/ci/print_revision_sql.sh#L14-L16

Not sure why they were needed...

Edit: Ah, probably because it gets passed into github output later... https://github.com/indico/newdle/blob/master/.github/workflows/migration-sql-comment.yml#L48
But it looks like the peter-evans/create-or-update-comment actions now supports body-path to point to a file instead of hardcoding a string, which should remove the need to use output altogether.

[tomasr8]

Thanks @renefs!!