Provide better visibility into migrations

[pdbossman]

It would be useful to have the following progress information during migrations:

1. Number for items (dynamodb/alternator) read
2. Number of items inserted in target (dynamodb/alternator)
3. If we implement discarding expired, output total items discarded

[julienrf]

AFAICT, there is no direct way to access such information.

There is a class AbstractReadManager which tracks the number of read items, but it is used internally only by emr-dynamodb-connector to adjust the read throughput, and this information is not publicly exposed. We would need to investigate further whether this class effectively tracks the total number of read items (and not just the amount of read items per partition). If that’s the case, a basic way to expose that information would be through logs. Otherwise, we would have to use a Spark accumulator to keep track of all the read items across all the partitions.