[Feature] Transaction isolation that supports the Serializable isolation level

### Search before asking

- [X] I searched in the [issues](https://github.com/apache/paimon/issues) and found nothing similar.

### Motivation

Currently, the transaction isolation level of paimon is snapshot level. For the following concurrent modification scenarios, incorrect data may occur

```scala
sql("create table t1 (id int, c1 string, c2 string) using paimon tblproperties('primary-key'='id')")

sql("insert into t1 values (1, 'a', 'a')")

val update1 = Future {
sql("update t1 set c1 = 'b' where id = 1")
}

val update2 = Future {
sql("update t1 set c2 = 'c' where id = 1")
}
Await.result(update1, 1000.seconds)
Await.result(update2, 1000.seconds)

sql("select * from t1").show()
```

The output result may be any of `1, 'b', 'a'` `1, 'a', 'c'` or `1, 'b', 'c'`

### Solution

_No response_

### Anything else?

_No response_

### Are you willing to submit a PR?

- [ ] I'm willing to submit a PR!

Start with the concurrent SQL example in the issue and trace Paimon's transaction-isolation implementation; no files, entry points, or tests are named. Done would mean a defined Serializable isolation level prevents the conflicting outcomes shown, with regression coverage.

Written by the indexing model from the issue text.