Process the shards using multiple processes in prepare_train_data

[hmashlah]

Process the shards using multiple processes in prepare_train_data.

I tested the change by running the following commands and checking the outputs

python -m sockeye.prepare_data --source source --target target --output prepared_data --max-processes 10 --min-num-shards 20

python -m sockeye.prepare_data --source source --target target --output prepared_data

Pull Request Checklist

• Changes are complete (if posting work-in-progress code, prefix your pull request title with '[WIP]'
  until you can check this box.
• Unit tests pass (pytest)
• Were system tests modified? If so did you run these at least 5 times to account for the variation across runs?
• System tests pass (pytest test/system)
• Passed code style checking (./style-check.sh)
• You have considered writing a test
• Updated major/minor version in sockeye/__init__.py. Major version bump if this is a backwards incompatible change.
• Updated CHANGELOG.md

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[fhieber]

Thanks, nice change! Only a few comments.

[fhieber]

Could you also bump the minor version (sockeye/__init__.py) and update the Changelog? Thank you!

[fhieber]

Thanks for iterating! How about making the max_processes flag in data_io.py a proper int with default 1 instead of None?

[fhieber]

Suggested change

	Load a shard source/target data files into an NDArrays and then save it to desk.
	Load shard source and target data files into NDArrays and save to disk.