abstract content, variant, and PIM entities

[Vandivier]

after we migrate from rect (#22 )

let's enhance the rect entities to include source, variant, and PIM concepts.

related #34

note: abstract content is an aggregator of variants: it is not published anywhere itself. variants are published over time and across channels. abstract content can have notes on it though, and we can also use it as a way to link related contents or even related abstract contents ("similar ideas" - manually identified as opposed to vector identification)

abstract content has title and public notes...so technically it could be possible to consume these by an LLM with no public variant consumption

[Vandivier]

admin notes should have public and private forms
public can be used for Material Search
private can include PII that is only for internal creator workflows (eg: "Jim is going to draft, Carol will approve, this is related to task foobar where Kim is leading, so email Kim at kim@foobie.dooby with questions")

[Vandivier]

note: by default

PIM has three forms (could be three totally different entities, or nah):

1. structured - using a Prisma model w various keys for notes, etc
2. stringified - a plain-text representation of the variant that is consumable by an LLM
3. cleaned - a plain-text representation of the variant that is output from an LLM

we should ultimately have 1 cleaned form of the abstract content per rect iteration (book, course, flash cards, quizzes)
by default, the abstract cleaned form === best variant cleaned form under the abstract (see #36)

we can also mark whole abstract content for inclusion or exclusion, or apply a performance filter on them, so that a variant will not be processed even if it is the best under an abstract but it is still not good enough for our rect-level bar. we can also include or exclude by tag, so that some very high performing videos might be excluded because they are comedy rather than instruction videos, for example.