Document combining lexical, dense, and sparse retrieval in production search

PR https://github.com/elastic/docs-content/pull/6727 adds a "When to use dense or sparse embeddings?" section to [Tutorial: Dense and sparse workflows using ingest pipelines](solutions/search/vector/dense-versus-sparse-ingest-pipelines.md). That page is intentionally structured as dense **versus** sparse (tab sets for model choice, mapping, ingest, and query).

During review, @leemthompo noted that production search often **combines** approaches rather than choosing only one embedding type-for example lexical (BM25/full-text) plus dense vectors plus sparse vectors (ELSER) in a tiered or multi-stage pipeline. That topic was left out of #6727 to avoid overloading a "versus" tutorial.

Add documentation on combining lexical, dense, and sparse retrieval in production search.