Version logic issue

[VladimirDVV]

Hello! I've encountered an issue with the _version column logic when switching replication modes and clearing the replicator's state.
I started the replication with skip_initial_replication: false to allow the tool to create the ClickHouse table structure and perform an initial sync.
During this process, the _version column reached a certain value (e.g., 100).
I stopped the process to manually add indexes to the ClickHouse table.
I cleared the contents of the binlog_replicator.data_dir folder to reset the internal state.
I changed the configuration to skip_initial_replication: true and restarted the replicator.
When the replication resumed in skip_initial mode, the new data coming from the MySQL binlog had _version values starting from 1.
Because I am using ReplacingMergeTree, ClickHouse considers the old records (with _version = 100) to be "newer" than the current real-time data (where _version restarted at 1). As a result, updates in ClickHouse are not being applied correctly because the "older" high-version records persist.

[bakwc]

For now added version_initial_value config parameter, you could set it manually to some bigger number if you removing state files and restarting replication. Later maybe add auto-detect of last version.