docs: gaps fixes from various sections

[prrao87]

Summary

Embeddings

• Clarify TypeScript provider registration before getRegistry().get(...).
• Document automatic embedding batch behavior: vector fields are generated only when absent or all-null.
• Add TypeScript string-query fallback behavior and first-embedding-function behavior for multi-embedding tables.

Indexing & Search

• Add FM index to the indexing overview and clarify scalar index defaults and list-column constraints.
• Document target_partition_size, IVF partition defaults, and fast_search() freshness caveats.
• Clarify that Lance-native FTS no longer accepts Tantivy-only options.
• Add default search limit, phrase-position requirement, and adaptive minimum_nprobes / maximum_nprobes behavior.

Namespaces

• Add list pagination and root namespace behavior using limit and page_token.

Reranking

• Add CrossEncoder local dependency and trust_remote_code details.
• Add LinearCombination fill, weight validation, and empty-side behavior.
• Add MRR _rowid requirement for multivector reranking.
• Add RRF K > 0 and TypeScript RRFReranker.create(...) note.
• Add evaluation prerequisite note for FTS index plus embedding metadata or explicit vectors.

Table Operations

• Add Python on_bad_vectors policy behavior for malformed vectors.
• Clarify that LanceDB primary keys are unenforced metadata used by operations like merge_insert, not write-time uniqueness constraints.

Validation

• git diff --check
• Added-line text check for excluded blob/Helm terms, duplicate adjacent additions, and very long added lines

Notes

• Did not update snippets because this patch adds concise behavior notes, not new canonical code examples.
• Did not patch the storage report item because the current Python sync source accepts storage_options.