refactor: modified RedisDataConn have a pool and RedisDataConn#get_connection get conn from pool

[sttk]

This PR modifies RedisDataConn#get_connection to retrieve a connection from r2d2::Pool<RedlisClient>, not to get a same connection instance that was created along with the RedisDataConn.

In the previous implementation, while the borrow of the conn: redis::Connection instance obtained from RedisDataConn#get_connection was active, you couldn't call the RedisDataConn#add_force_back method. To perform multiple operations and register force-back functions alternately, you had to enclose the conn acquisition and operations in a block. After the block ended and the conn was dropped, you could then execute add_force_back, and then repeat the process of acquiring the conn and performing operations inside another block.
In the new implementation, it's possible to execute operations on the conn and add_force_back alternately.

Also, in the previous implementation, it was impossible to handle a connection that was severed while being passed around. With the new implementation, however, the pool automatically reconnects, allowing you to get an active connection by calling RedisDataConn::get_connection.
However, if all connections are in use, there will be an overhead of waiting until a connection is released.

[Copilot]

Pull Request Overview

This PR refactors the RedisDataConn struct to use a connection pool instead of holding a single pooled connection, improving concurrent access and connection management.

• Replaces direct connection storage with r2d2::Pool<redis::Client> reference
• Modifies get_connection to return a pooled connection on-demand rather than a mutable reference
• Updates force_back method to acquire fresh connections from the pool for rollback operations

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}