Add streaming query API with metadata tracking support

[bplatz]

Summary

Adds streaming query API that emits individual results as they are produced, rather than collecting all results into a vector. Supports all query types (SELECT, SELECT ONE, CONSTRUCT) and formats (FQL, SPARQL).

Key Features

• Memory efficient streaming: Results emitted individually for large result sets
• Metadata tracking: Optional :meta tracking emits final metadata map after results
• CONSTRUCT streaming: Emits ungrouped graph nodes without @graph wrapper
• Full query support: ORDER BY, GROUP BY, and aggregates (collect before emitting)
• Cache validation: Early validation rejects :cache option for streaming queries

API Functions

• fluree.db.api/query-stream - Stream query against database
• fluree.db.api/query-connection-stream - Stream query via connection

Metadata Format

When :meta true is enabled, final result is:

{:_fluree-meta {:fuel 123, :time "5ms", :policy {...}, :status 200}}

Test Coverage

10 tests, 39 assertions covering:

• Basic streaming (SELECT, LIMIT, SELECT ONE)
• Metadata tracking (FQL and SPARQL)
• Connection-based queries
• CONSTRUCT queries (FQL and SPARQL)

Implementation Notes

• Cache validation moved to public API layer
• Internal functions simplified (validation upstream)
• CONSTRUCT uses cat transducer for streaming ungrouped nodes
• Metadata calculated from tracker counters without forcing realization

[zonotope]

@bplatz it looks like this pull request is based on the bm25 ns branch, but the base branch is set to main so there are a lot of unrelated changes here.

Could you update the base branch to make reviewing easier?

[bplatz]

Closed in favor of #1186