[WIP] Schema and InternalModel improvements to reduce overhead

[runspired]

Some notes:

• It seems many of ember-data's tests are intricately tied to specific implementations of the Internals even when not needed for the test in question. We need to funnel test abuse of private API through a central location so we can address these errors more quickly and refactor more easily.
• It seems that many of the adapter and serializer methods are wrapped in multiple abstraction layers that make refactoring harder, likely carry perf costs, and obfuscate code patterns.
• This change does not actually do the batching that we should do, it just unlocks being able to do so.
• Need to update heimdall-query to show counter stats
• Need to update heimdall-instrumentation PR with counter stats, docs, feedback
• @bcardarella was so kind as to point me to this thing he made, should be updated to show mixin use as well but it begins to reveal some of the complexity that needs to be flattened. https://ember-twiddle.com/#/8343688ec39c38b5cb4dc4c29f76905a?fullScreen=true
• perf benchmarks: While this is still very WIP, here are preliminary perf results (~10% improvement).

Baseline:

With Changes: