Fix: nested recursive relationships consumed by first reference

Fix: nested relationship preload on cache hit

Cache was previously used to just prevent infinite recursion if a
relationship type was already processed. This would prevent preloading
of a type if it was previously encountered, even for indirect relationships.

Direct relationship example (tags excluded for brevity):

type A struct {
	ID int
	A []A
	AB []AB
}

type B struct {
	ID int
	AB []AB
}

type AB struct {
	A A
	B B
}

Valid: query.A.A.AB.B
Invalid: query.A.AB.B --> query.A.A has already "consumed" the child relationships

Instead, use cache for memoization and store previously computed values,
so that it can be reused.

fix: Fix Query field names

fix: Client use exported interfaces

fix: Interface decomp to avoid method shadowing (i.e., underlyingDB a…

…nd UnderlyingDB)

fix: Fix invalid struct template