[Refactor]: Unify the vector, text, sparse, and tensor search parameter style

### Describe the improvement you'd like

TopN in sparse, vector are mandatory and digit, which is key-value style option in tensor and text search. We need to unify the style.

Start by locating the parameter definitions and call sites for vector, sparse, tensor, and text search. Compare how TopN and digit are represented, then determine the unified style and verify that all four search modes use it consistently.

Written by the indexing model from the issue text.