HOTFIX: A few PySpark tests were not actually run

This is a hot fix for the hot fix in fb499be1ac935b6f91046ec8ff23ac1267c82342. The changes in that commit did not actually cause the `doctest` module in python to be loaded for the following tests:
- pyspark/broadcast.py
- pyspark/accumulators.py
- pyspark/serializers.py

(@pwendell I might have told you the wrong thing)

Author: Andrew Or <andrewor14@gmail.com>

Closes #1053 from andrewor14/python-test-fix and squashes the following commits:

d2e5401 [Andrew Or] Explain why these tests are handled differently
0bd6fdd [Andrew Or] Fix 3 pyspark tests not being invoked

1 files changed, 4 insertions, 1 deletions

diff --git a/python/run-tests b/python/run-tests
index 3b4501178c..9282aa47e8 100755
--- a/python/run-tests
+++ b/python/run-tests
@@ -44,7 +44,6 @@ function run_test() {
         echo -en "\033[0m"  # No color
         exit -1
     fi
-
 }
 
 echo "Running PySpark tests. Output is in python/unit-tests.log."
@@ -55,9 +54,13 @@ run_test "pyspark/conf.py"
 if [ -n "$_RUN_SQL_TESTS" ]; then
   run_test "pyspark/sql.py"
 fi
+# These tests are included in the module-level docs, and so must
+# be handled on a higher level rather than within the python file.
+export PYSPARK_DOC_TEST=1
 run_test "pyspark/broadcast.py"
 run_test "pyspark/accumulators.py"
 run_test "pyspark/serializers.py"
+unset PYSPARK_DOC_TEST
 run_test "pyspark/tests.py"
 run_test "pyspark/mllib/_common.py"
 run_test "pyspark/mllib/classification.py"