Add TopK to Cloud Service

[davidmyriel]

Adds one entry for TopK to the Cloud Service section.

- [TopK - Managed search engine for accuracy-critical AI applications, combining hybrid search, multi-vector (late-interaction) retrieval, custom ranking, and managed inference in a single API](https://topk.io)

Formatted to match the section's [Name - description](url) convention.

What it is: a managed search engine aimed at retrieval quality rather than vector storage alone. Hybrid dense plus BM25 scoring, multi-vector retrieval with MaxSim for late-interaction models such as ColBERT and ColPali, user-defined ranking expressions over vector scores, BM25 scores, and metadata, plus embedding and reranking models hosted alongside the index.

Why it is not a duplicate of what is already listed: the existing Cloud Service entries are largely managed vector stores. TopK indexes multi-vector fields so MaxSim retrieves candidates directly, rather than rescoring a shortlist produced by a single-vector or keyword query.

Source and SDKs are at https://github.com/topk-io/topk (Rust, MIT, actively developed), docs at https://docs.topk.io.

Note the name collides with top-k, the ranking concept, so the description leads with what the service does to keep them apart for readers scanning the list.