[MINOR] Enable PySpark SQL readerwriter and window tests

[JoshRosen]

PySpark SQL's readerwriter and window doctests weren't being run by our test runner script; this patch re-enables them.

[rxin]

I wonder if we should create a simple linter rule in lint-python to check for stuff like this ...

[rxin]

or -- even better, should run-tests just run every python file?

[JoshRosen]

In principle, we should probably be using a proper testrunner like nose to handle test discovery and execution. The reason why we didn't do this initially is because we need some custom code in __main__ to emulate a shared SparkContext fixture for the doctests, since putting SparkContext setup and teardown code into each doctest would be very messy and slow.

In the medium-term, we're going to want to refactor run-tests anyways in order to make it easier to run subsets of the tests and python versions (see #4269, for example).

[JoshRosen]

Integrating my coverage reporting harness into the Jenkins builds would also help to catch this problem, since that makes it really obvious when test code isn't run.

[SparkQA]

Test build #33864 has finished for PR 6542 at commit 9f46ce4.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[rxin]

Merging in master & branch-1.4.