Exclude unloadable files at rsync time

Changed video-timeline workflow to use user_id instead of username.

Copy mysql database to bigquery dataset

* Don't include files starting with underscore.
* Switch away from dynamic dependencies for BQ load.
* Delete marker table on overwrite, at initialization.
* Add flag to skip clear_marker calls on overwrite.
* Add filtering parameter: exclude_field list.
* Always pass column names to sqoop.
* Add table-level descriptions.
* Add warehouse_subdirectory.
* Handle tinyint(N) where N>1, and output BIT as a STRING.
* Add comment on how blobs-to-BYTES doesn't work.
* Add basic unit tests.

Create an empty 'course_program_metadata' table for insights.

Don't include DownstreamMixin in "Table" Tasks.

And also remove from "Partition" tasks that don't need it.

Goal is to be able to specify overwrite_n_days on the command-line,
and not require it to have to be in config files.

Update args in running-tasks doc to use hyphens.

Make interval_start optional

Make it required only when interval is not specified.

Delete manifest file on overwrite.

Implement for module engagement to begin with, since its
ModuleEngagementDataTask class will often have the same
parameters (i.e. the given day to calculate), and therefore
uses the same manifest file when being rerun.

Renames TimestampPartitionMixin.date to TimestampPartitionMixin.datetime

Add quality check to travis builds.

Fix basic quality so it passes.

ENT-777 Pipeline tasks to import enterprise enrollment data

to better reflect that parameter's type.

Also moves the partition_value testing into test_course_list, since this is
where the TimestampPartitionMixin lives.

Use Luigi return codes.

Update Acceptence tests for data czar export key changes

Doc for JSON list input and incremental video.

Doc generation requires a single requirements file

When writing to S3, don't validate the bucket.

If Boto is passed "validate=True", it will require an additional
permission to be present when asked to list all of the keys in the
bucket.  We want to minimize the set of required permissions so we get a
reference to the bucket without validating that it exists.  It should
only require PutObject and PutObjectAcl permissions in order to write to
the target bucket.

This is particularly the case with answer-distribution, which writes out
to a bucket that is used by LMS for course downloads.

Upgrade to fork of Luigi 2.7.

Hive partition overwrite by default

Generate the Hive Table but not Table Data