docs: providing pre-generated document embeddings for auto-embedding fields

[alangmartini]

What

Adds a section to the 30.2 vector search docs covering a supported but previously undocumented pattern: generating document embeddings outside Typesense while keeping an auto-embedding field, so Typesense still generates the query embedding at search time and semantic/hybrid search works unchanged.

When a document already contains a value for its auto-embedding field, Typesense indexes the provided vector as-is and skips generation (see Index::batch_embed_fields, which skips docs that already carry the field on create/upsert/update).

Why

This is the pattern we recommend to customers doing large bulk imports alongside highly concurrent hybrid search: with built-in models, document embedding and query embedding share the same per-node model instance (and GPU), so import-time embedding can slow query embedding. Until now the docs only offered fully external embeddings with a manual vector_query (embedding the query client-side), so there was nothing to link customers to for this middle path.

Changes

• New "Providing pre-generated document embeddings for auto-embedding fields" section at the end of Option B: how it works, when to use it, an import example, a warning that vectors must come from the same model with the model's indexing prefix applied (e.g. passage: for ts/e5-*), and a note that updating embed.from fields without the vector triggers regeneration.
• A tip in the Hybrid Search section cross-referencing the new section as an alternative to manually embedding the query.

Only 30.2 is touched, per the convention of editing the latest version dir.