chore: Change EventInsertable to store the parsed Event object instead of its raw blocks

[stbrody]

builds on #481

[stbrody]

I hate this. Encoding it to a CAR just to re-encode it to the raw blocks is doing two passes over the data for no reason. I tried to add a method to unvalidated::Event that worked just like encode_car() but would instead return a Vec<EventBlockRaw> so we could get what we need in a single pass. The problem is that the EventBlockRaw type lives in the ceramic-store crate while the unvalidated::Event type lives in the ceramic-event crate, but the store crate depends on the event crate, so I can't make the event crate depend on the store crate without introducing a circular dependency.

Would love any suggestions on how to clean this up.

[stbrody]

This is passing all tests except the migration tests, I might need a hand to figure out what's up with the migration tests.

[stbrody]

Okay I figured out a part of what's wrong with the migration tests - there are some event CAR files being used in that test that don't round-trip through unvalidated::Event. Pushed a change that demonstrates that. Now the question is: is this a bug in our serialization/deserialization logic? Or is the migration test using invalid CAR files?

@nathanielc