[SPARK-8541] [PYSPARK] test the absolute error in approx doctests

A minor change but one which is (presumably) visible on the public api docs webpage.

Author: Scott Taylor <github@megatron.me.uk>

Closes apache#6942 from megatron-me-uk/patch-3 and squashes the following commits:

fbed000 [Scott Taylor] test the absolute error in approx doctests

(cherry picked from commit f0dcbe8)
Signed-off-by: Josh Rosen <joshrosen@databricks.com>

python/pyspark/rdd.py

@@ -2198,7 +2198,7 @@ def sumApprox(self, timeout, confidence=0.95):
 
         >>> rdd = sc.parallelize(range(1000), 10)
         >>> r = sum(range(1000))
-        >>> (rdd.sumApprox(1000) - r) / r < 0.05
+        >>> abs(rdd.sumApprox(1000) - r) / r < 0.05
         True
         """
         jrdd = self.mapPartitions(lambda it: [float(sum(it))])._to_java_object_rdd()
@@ -2215,7 +2215,7 @@ def meanApprox(self, timeout, confidence=0.95):
 
         >>> rdd = sc.parallelize(range(1000), 10)
         >>> r = sum(range(1000)) / 1000.0
-        >>> (rdd.meanApprox(1000) - r) / r < 0.05
+        >>> abs(rdd.meanApprox(1000) - r) / r < 0.05
         True
         """
         jrdd = self.map(float)._to_java_object_rdd()