Introduce record normalizers

[dgeb]

This PR introduces the RecordNormalizer interface and StandardRecordNormalizer implementation. Normalizers are used to transform records and identities into the normalized form from a more developer-friendly form. They will be used by query and transform builders to check and transform inputs. Because normalizers are provided with a keyMap and schema, they can provide validation and also look up identities from keys.

A query builder can now accept either a RecordIdentity or RecordKeyValue in place of an identity. This makes the following queries possible:

source.query(q => q.findRecord({ type: 'planet', id: '123' }));
source.query(q => q.findRecord({ type: 'planet', key: 'remoteId', value: '1' }));

source.query(q => q.findRecords([{ type: 'planet', id: '123' }, { type: 'planet', id: '124' }]));
source.query(q => q.findRecords([{ type: 'planet', key: 'remoteId', value: '1' }, { type: 'planet', key: 'remoteId', value: '2' }]));

Normalizers can work with any kind of inputs, including model classes, such as in ember-orbit. This will soon unlock the ability to mix models with builders, such as:

source.query(q => q.findRecords([earth, jupiter]));

In order to support record normalizers, this PR also:

• Introduces [Async/Sync]Queryable and[Async/Sync]Updatable interfaces that standardize the public facing interfaces shared by sources and caches.

• Updates all record-based caches and sources to expose builders as generics. Individual implementations can then override a builder with a custom normalizer and have typings still work.

• Fixes a number of test cases that were using improperly normalized data.

• Deprecates the no longer needed firstResult fn in @orbit/utils.

---

Fixes #545
Fixes #547