Automate uploads and amendments

[hancush]

Description

This PR:

• Containerizes data making
• Migrates to our contemporary strategy for running migrations in the entrypoint
• Unifies the upload and amendment data pipelines
• Adds gut checks to intermediate targets where data can be lost
• Adds a management command to facilitate uploading new data or amending existing data
• Adds targets to upload or amend a given file using the management command
• Fixes two import bugs
  • Employer names can be changed via the admin interface, but a corresponding alias is not created for the new name. If the updated name appears in a new upload, the import tries to insert it as a new employer, violating the unique name constraint. This PR adds a data migration to insert aliases based on Django admin log entries indicating a name change, and adds logic to the employer admin to create aliases on name change automatically, moving forward.
  • When adding a third year, our query to link people failed when a person already had two salaries, because having more than one salary caused a person to appear twice in our query selecting existing people, and we only link if there is one matching person. I added a distinct on person ID to the query to resolve this issue.

Connects #248.

Testing instructions

Tested locally.

[hancush]

@fgregg This is ready for preliminary review. I would especially appreciate feedback on overall approach and thoughts on high value tests to write, if any.

[fgregg]

I'm tempted to write a little spec in marshmallow and use that. How long would that take you @hancush ? Do you think that could be of value?

[hancush]

TODO

• [x] Set CELERY_RESULT_BACKEND = REDIS_URL in deployment settings
• [x] Debug failures to upload to S3. Getting...

# info about PUT request
encountered a retry exception: Saw RequestTimeout, retrying
establishing HTTPS connection: host=bga-payroll-database.s3.amazonaws.com, kwargs={'timeout': 70, 'port': 443}