perf: batch embedding calls in MaximalMarginalRelevance (2N → 1 call) [stacks on #2496]

[pidefrem]

What does this PR do?

perf: batch embedding calls in MaximalMarginalRelevance (2N → 1 call)

MaximalMarginalRelevance.extract_topics() calls the embedding model twice per topic inside a loop: once for the individual candidate words and once for the concatenated sentence. With N topics, this means 2N separate embedding calls — each with its own model inference overhead, GPU kernel launch, or API round-trip.

Changes:

Collect all candidate words and all topic sentences across all topics into a single flat list. Make 1 embedding call for the combined list. Slice the result array back into per-topic chunks using pre-computed index ranges.

This reduces 2N calls to exactly 1 call, with identical output. The improvement scales linearly with topic count and is most impactful with API-based embedding models (e.g., OpenAI: 100 API calls → 1 for 50 topics).

Benchmark (to be filled in before submission):

Setup (N topics, sim. latency)	Embedding calls before	after	Wall time before	after
50 topics, 0 ms	2N (TODO)	1 (TODO)	(TODO)	(TODO)
50 topics, 10 ms/call	2N (TODO)	1 (TODO)	(TODO)	(TODO)

The call-count reduction (2N → 1) is deterministic and is the core argument; the wall-time row illustrates the real-world impact with remote embedding models.

Fixes #2498

Before submitting

• This PR fixes a typo or improves the docs (if yes, ignore all other checks!).
• Did you read the contributor guideline?
• Was this discussed/approved via a Github issue? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes (if applicable)?
• Did you write any new necessary tests?

---

⚠️ Draft — stacks on #2496. This branch includes the _extract_representative_docs fix commits from #2496 as a prerequisite (marked [prereq]), so the diff here shows that work too. Once #2496 merges, I'll rebase onto master and the prereq commits drop out, leaving only the MMR batching change. Marking ready for review after #2496 lands.