refactor: grant_ownership retry logic

[TCeason]

I hereby agree to the terms of the CLA available at: https://docs.databend.com/dev/policies/cla/

Summary

Before this pr, if the transaction execution of granting ownership fails, it will be retried immediately.

This often leads to frequent failures when multiple connections are simultaneously executing grant ownership.

This is because the role key may cause conflicts. A better solution is to store object and privilege separately as key-value pairs.

However, this solution involves significant changes. This PR temporarily adjusts the retry mechanism(
Tthe role information will be reacquired with each retry, aiming to minimize conflicts.).

No TxnRetryMaxTimes errors occurred during local testing.

Local Test

Environment

create user a identified by '123' with DEFAULT_ROLE='role1';
create role role1;
grant role role1 to a;
grant create database on *.* to role role1;
-- user a
create database a;

script1:

$ cat create.sh
#!/bin/bash  
  
for((i=1;i<=5000;i++));  
do   
	echo "create table if not exists a.t_$i(id int)" | bendsql -ua -p123

done 

script2:

$ cat create1.sh
#!/bin/bash  
  
for((i=5001;i<=10000;i++));  
do   
	echo "create table if not exists a.t_$i(id int)" | bendsql -ua -p123

done 

Test

-- session1
bash ./create.sh

-- session2
bash ./create1.sh

Error Error: APIError: ResponseError with 2317: TxnRetryMaxTimes: Txn grant_ownership has retry 60 times, abort. not occurs.

• Fixes #[Link the issue here]

Tests

• No Test refactor while logic. if all ci pass, pr is ok.

Type of change

• Refactoring

---

This change is 

[drmingdrmer]

Reviewed 1 of 1 files at r1, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @BohuTANG)