Chuck/dev/reset schema

[chuckadams]

Pull Request

What changed?

• Migrations have all been rolled into a minimal set. This necessitates a db reset with make init, which means you'll need to regenerate API keys (I'll back up and restore the users and tokens on prod to minimize disruption)
• Migrations now assume Postgres, in that all varchar columns are now text columns without limits (which is the only kind of varchar postgres has anyway). SQLite could probably cope with this schema too, but not when I start adding support for trigrams, so unless until we need sqlite support, I'm leaving it at pg-only.
• all jsonb columns except ac_raw_metadata are removed, attributes have been added to Plugin and Theme to make the former columns computed from the raw column. This should be transparent to most uses.
• As a side effect of using computed attributes, all url rewriting is now done on the fly in the accessors. I'm pondering moving the rewrite logic elsewhere, but it can stay there for the time being.

Why did it change?

The meandering saga of our database migrations includes the AspireSync tables and various missteps and corrections. I'd like something I can easily search with my IDE without the results being full of red herrings.

Normalizing the tables to use the computed attributes makes things like PluginProps and ThemeProps smaller and simpler. It does remove the ability to declare every last bit of plugin state in the constructor, but we can still parse metadata the same way as before.

Did you fix any specific issues?

closes: #125
closes: #126

CERTIFICATION

By opening this pull request, I do agree to abide by the Code of Conduct and be bound by the terms of the Contribution Guidelines in effect on the date and time of my contribution as proven by the revision information in GitHub.