improve use of LLJIT

Current status
- every SQL construct standalone JIT instance, destroyed after single SQL query finished

Better ?
- reuse JIT instance from multiple SQLs
- remove module with ResourceTracker

The issue concerns the LLJIT usage in OpenMLDB's SQL execution. Look for code that creates and destroys JIT instances per query, likely in the SQL execution engine or compiler components. Research how LLJIT instances and ResourceTrackers are currently managed. The goal is to modify the lifecycle to allow reuse across multiple queries, which requires understanding the project's JIT architecture and threading model.

索引モデルが issue の本文から書いたものです。