S3: API now also returns number of bytes processed

[bblommers]

Also simplifies the flow. The input is now processed one document at a time:

for document in source:
    # process FROM clause
    # check if WHERE clause is applicable
    # process SELECT clause
    add document

Which should make it much easier to add support for a LIMIT-clause later on.

Note that only the S3-part has been changed, the DynamoDB part is unchanged.