Optimization for DDL execution

Feature Request

Is your feature request related to a problem? Please describe:

In the current replication model, DDL is executed synchronously and blocks all DMLs in other tables. This could cause replication block in some scenarios

• the DDL is slow, such add index to a large table.
• the replication of one table is slow, other tables will have to wait for this table catch up because the DDL is a strong barrier.

Describe the feature you'd like:

• Currently DDL execution has no timeout, should we add some restriction. https://github.com/pingcap/ticdc/pull/936
• Once the add index command has been sent to TiDB, even the client is timeout, we can ignore this error and continue replication.
• Could we add some mechanism to allow single table DDL doesn't block DML replication of other tables.