Tighten up the deployment process

[babolivier]

There aren't a lot of guard rails for deploying ISPDB changes to production at the moment. A single push to the prod branch triggers it, and there isn't a mandatory flow that dictates how this push should happen. Similarly, this means we don't have any enforcement in place around using the staging environment to identify issues before they make their way into production.

I'm thinking of tightening up the deployment process for production. This would involve:

• adding a CI step for PRs to prod that runs convert.py, checks it doesn't fail, and compares its output to what's currently on the staging environment
• protecting the prod branch so it cannot directly be pushed on
• making the aforementioned CI step mandatory to be able to merge PRs into prod

This means requiring a PR and running through the CI (including extra prod-specific steps) before any change can be deployed in production.