Single-read pagination extractor (items + cursor in one pass)

[OmarAlJarrah]

Problem

CursorPaginationStrategy takes two (Response) -> … lambdas that each read a single-use body; a test already works around the double-drain with an IdentityHashMap cache (CursorPaginationTest.kt:47). That's a correctness trap in a public API.

Proposed change

Replace the two body-reading lambdas with one extractor returning items + next cursor together. After the fix, delete the test's caching workaround as proof.

Acceptance

• One body read per page
• Test workaround removed
• apiDump

Dependencies

• Blocked by Remove TokenPaginationStrategy (duplicate of CursorPaginationStrategy) #39 — remove TokenPaginationStrategy first
• Related Unify the two pagination stacks into one #30 — pagination-stack unification

---

Priority: medium · Effort: medium