[FLINK-37313] Fix the problem of reading binlog before the high and low watermarks during the snapshot process

[gongzexin]

This closes FLINK-37313.

SnapshotSplitReader#isBackfillRequired

If the binlog has not changed during the snapshot split read, the binlog should not be read.

MySqlStreamingChangeEventSource#handleChange

Even if you read the binlog of a table you don't care about, you should update the timestamp to avoid MySqlBinlogSplitReadTask#handleEvent reading a large amount of binlog

[lvyanquan]

Hi, @gongzexin.

I think that the problem that you met was resolved in #3902 as #3902 (comment) describes, can you check for this?

[gongzexin]

Hi, @gongzexin.

I think that the problem that you met was resolved in #3902 as #3902 (comment) describes, can you check for this?

Hi, @lvyanquan.
The modification of BinlogOffset.java in the above PR does indirectly fix the problem of being hung when reading binlogs between high and low watermarks during snapshot. However, there seems to be no check on whether the binlogs of high and low watermarks should be read during snapshot. Do you know about this problem?