ChannelToken migration

[Bjvanminnen]

First in a series of migrations around ChannelTokens. Overarching plan is as follows

1. [migration] Create new (indexed) column in channel_tokens called storage_id
2. [code change] Update code to write both user_id and storage_id columns for new rows
3. [migration] Backfill storage_id, derived by decrypting channel column. On production this will be managed via a script that batches updates. In other envs this will involve a migration.
4. [migration] Make storage_id non-nullable, make user_id/channel nullable (now that all rows should be backfilled and new rows are getting storage_id from 2
5. [code change] Stop using channel/user_id, use storage_id instead
6. [migration] Delete channel/user_id columns

[jopolsky]

Can you include the query changes to validate the index?

[Bjvanminnen]

The query changes are going to be that eventually both here
https://github.com/code-dot-org/code-dot-org/blob/staging/dashboard/app/models/channel_token.rb#L34
and here
https://github.com/code-dot-org/code-dot-org/blob/staging/dashboard/app/models/channel_token.rb#L43

We search using storage_id instead of user_id.

[davidsbailey]

This looks like the right plan to me.

[Bjvanminnen]

For posterity, the migration step of the DTP that contained this PR took ~3 minutes (based on slack logs)