Persist edit history for site configuration including author and expose via GraphQL

[mrnugget]

Part of https://github.com/sourcegraph/sourcegraph/issues/40377 but only about site-configuration.

Tasks

• Persist author for site config changes (we already persist the history in critical_and_site_config but don't record the author) internal: Persist author in site config #46150
• Maybe, depending on whether it's encrypted: re-encrypt history when encryption key changes?
  • We do not encrypt site config at the moment in the application layer
• Define API to expose the site config history via GraphQL graphqlbackend: Add API spec for site config history diff #46205,
  • API would need to expose history with date/time, who, diff to previous edit.
  • Needs to support pagination
  • Draft:

    siteConfiguration {
      // ...
      editHistory(last: 10) {
      nodes {
        id
        author { firstname, lastname } // <-- this is a User node
        date
        diff // <-- this is a diff
        // lines changed? removed? we already have a "diff" node in our GraphQL API, maybe we can reuse?
      }
      pageInfo {
        totalCount
        // ...
      }
      }
    }

• Implement API (Fully implemented in: https://github.com/sourcegraph/sourcegraph/pull/47310)
  • Handle changes to "redacted" values
  • Compute diffs on the backend after loading requested number of edits

[indradhanush]

Had a chat with @mrnugget today and here's the plan going forward for generating diffs which is yet to be implemented:

Database changes

• Add a new column to the critical_and_site_config table called redacted_contents Store redacted site config in DB #46781
• On write, store the redacted version of the config to this column but:
  i. Before redacting, replace each secret in the site config with its sha256 hash modulo 1000 and prefix it with REDACTED-, such that the access token would now look like: REDACTED-999 for example Store redacted site config in DB #46781

Diff implementation

• When generating the diff, look at redacted_contents which will now pick up the diff on access tokens since underlying hash will have changed if the access token was updated internal/database: Scan redacted_contents from critical_and_site_config #47079 graphqlbackend: Implement diff and reproducedDiff #47310

API changes

• Add a new boolean field to SiteConfigurationChange named reproducedDiff which indicates if the diff can be generated or not graphqlbackend: Implement diff and reproducedDiff #47310
  • For site config entries prior to introduction of this feature, it is not possible to generate the diffs safely without leaking the access token themselves
• Make Diff nullable so that for changes with no diff it returns null instead of an empty string graphqlbackend: Implement diff and reproducedDiff #47310