Maestro should require channel classifications to be a specific set of values

- [ ] This issue is blocking
- [ ] This issue is causing unreasonable pain

Currently, when creating a channel, we can specify any string as the classification for the channel. Mostly we use "test", "dev", and "product". There are parts of the validation and release infrastructure that do different things depending on the channel category. Instead of allowing any value, We should treat these categories as a bounded set so that the decision of which classification to use is better understood and we avoid potential errors. Some examples could be:

* engineering
* tooling
* product
* ?

No files, tests, or entry points are named. Start by locating channel creation and classification validation, then resolve the canonical bounded set with maintainers; done means invalid classifications are rejected consistently and existing validation and release behavior remains correct.

Written by the indexing model from the issue text.