source: better handling of headers of type Data

[lburgazzoli]

In the CamelSourceTask, if a Camel's header is of type date, it gets propagated as a string to Kafka with something like:

if (value instanceof Date) {
    record.headers().addString(
        keyCamelHeader, 
        new SimpleDateFormat("yyyy-MM-dd").format(value));
}

this approach has some issues:

1. the consumer is not able to know the real type of the header as it is just a string
2. we lose precision as we remove the time of the day

As Kafka support date and time with org.apache.kafka.connect.data.[Data|Time], we should probably avoid to use a string to represent dates however to fully translate a java.util.Date to the related Kafka types, we'd need to split it to multiple headers keys or using multiple headers for the same value.

I don't have enough knowledge at the moment about what could be the best solution but at least we should start by not losing the type.

[lburgazzoli]

It turn out that it it much simpler as there is org.apache.kafka.connect.data.Timestamp that represent a date.
However we should remove support for java.sql.Date and java.sql.Timestamp