There is a lot of value validation that happens in the *MetricType classes, such as truncating strings when they are too long. In some cases for technical reasons, the validation is split between the *MetricType and *StorageEngine classes (individual strings in a string list vs. the string list as a whole).

All of this should be moved to the *StorageEngine classes. There are at least two good reasons for this:

1. When we oxidize glean, we will be oxidizing the internal classes (*StorageEngine) not the public API interface classes (*MetricType). Having important logic in the *MetricType classes means we will need to duplicate that logic across all flavors of glean (Android, iOS etc.)

2. Notating errors currently has to happen in a separate coroutine worker from the recording of the metric value itself. This causes race conditions that makes testing harder and requires the addition of a "hack" to record errors synchronously during testing.

This was merged \o/ closing it!