Test the performance and memory usage of sdk when OpenMLDB has many tables(1w+)

**Is your feature request related to a problem? Please describe.**
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

RTIDB java client will use G-level memory when RTIDB has 1w+ tables. So we need to test OpenMLDB in this scene.

**Describe the solution you'd like**
A clear and concise description of what you want to happen.

**Describe alternatives you've considered**
A clear and concise description of any alternative solutions or features you've considered.

**Additional context**
Add any other context or screenshots about the feature request here.

The issue mentions testing the OpenMLDB SDK's performance and memory usage with over 10,000 tables, referencing a similar issue with RTIDB Java client. Start by examining the OpenMLDB SDK codebase, particularly the client-side libraries and any existing performance tests. Look for benchmarks or stress test scripts. Determine how to create and manage many tables programmatically, then measure memory consumption and performance metrics. The goal is to produce a test report or script that can be run to validate the SDK under high table counts.

索引モデルが issue の本文から書いたものです。