feat(voyageai): refresh models, add contextualized embeddings + token-aware batching

[fzowl]

What

Refreshes the Voyage AI integration to the current model catalog and adds first-class support for the contextualized-chunk embedding models.

Embeddings (kotaemon/embeddings/voyageai.py)

• Contextualized models (voyage-context-4, voyage-context-3) via the contextualized_embed API. Each input string is embedded as its own independent document: the batch is sent as a flat list[str] with enable_auto_chunking=True and chunk_size=32000, so every string resolves to exactly one chunk and one embedding. This yields deterministic per-input vectors and trivial result collection. Cross-input contextualization (inputs=[batch]) is intentionally not used, because generic embed_many callers pass unrelated texts. This is documented on the class.
• Query path: the API rejects auto-chunking when input_type="query", so the retrieval path disables it — enable_auto_chunking = input_type != "query", and chunk_size is dropped when disabled.
• Token-aware batching for the plain-text path (embed/embed_many and async equivalents): batches are bounded by both the item count (MAX_BATCH_SIZE) and the per-model total-token budget, not only a fixed item count. A single oversized text is still sent on its own.
• Fixes the async path, which previously awaited an attribute instead of the coroutine.

Rerankings

• Default updated to the current rerank-2.5 model; help text lists the current models.

Config

• VOYAGE_RERANK_MODEL env var wired into flowsettings.py; model options documented in .env.example.

Tests

• Contextualized document path: flat list[str], enable_auto_chunking=True, chunk_size=32000.
• Contextualized query path: auto-chunking disabled, no chunk_size.
• Batching logic: split at the token boundary, a single oversized text goes through alone, and the item-count cap is respected.

Validation

The repo's Python environment has a pre-existing langchain/langchain-core version mismatch that blocks collecting the kotaemon test suite (unrelated to this change). Ran the affected tests in an isolated environment with compatible langchain pins: all 6 Voyage AI tests pass; the only unrelated failure is the pre-existing Cohere embedding test. flake8 (max-line-length 88), black, and isort are clean on the changed files.

[fzowl]

CI was failing because the lockfile pinned voyageai==0.3.2, whose Client predates the contextualized_embed API — the two contextualized-embedding tests errored at patch time and the feature could not work at runtime. I've pushed a fix:

• Require voyageai>=0.4.0 (the version that adds the flat-list contextualized_embed with enable_auto_chunking/chunk_size) and relocked to 0.5.0.
• On Python 3.13 voyageai>=0.4.0 wants numpy>=2.1, which conflicts with unstructured<0.16 (numpy<2), so a marker keeps the older voyageai there; the test matrix (3.10/3.11) gets 0.5.0.
• voyageai declares langchain-text-splitters>=0.3.8, which would otherwise cascade a langchain-core 0.3 bump across the app. It only needs RecursiveCharacterTextSplitter (present since 0.2.x) for the optional local chunk_fn, never invoked on the server-side auto-chunking path, so a uv override-dependencies keeps the splitter on the 0.2 line and leaves the langchain stack untouched.

Only voyageai changes version in uv.lock; the rest of that diff is marker re-serialization from the 3.13 split. Unit tests on 3.10 and 3.11 are green.