No need for the logs on that level.

This allows Ghostferry to be interrupted.

The library gained the ability to resume the BinlogStreamer and the
DataIterator from arbitrary states. We're still missing the
reconciliation process to make the interrupt/resume behaviour correct
with respect to data safety. We're also missing the ability to interrupt
from within the IterativeVerifier.

Additionally, this patch exposes the need to refactor how we initialize
DataIterator/IterativeVerifiers, and the other "public" components that
are below the level of Ferry. It is currently difficult to initialize
these structs without Ferry and causes things like the hacks within
RunOnlyDataCopyForTables (previously named RunStandaloneDataCopy). We
can get rid of this in a future PR.

Made BinlogStreamer independent of Ferry.Config

Previously BinlogStreamer was dependant on Ferry.Config. It even went so
far to change Ferry.Config for the automatic MyServerId generation. This
will become problematic as we create new BinlogStreamer instances to
monitor the target database to detect schema changes.

We also exposed methods to create and initialize DataIterator,
BinlogStreamer, and all the major components from Ferry, so another
object that operates at the same level as the Ferry can be created (for
example: the IterativeVerifier, the DeltaCopier, the Reconciler).

The StateTracker was also made optional from the previous commit, as
this does make things kind of easier to work with.

If you want to resume with mismatched version, just manually change the
JSON.