IMP add a dataframe breakpoint method for testing

Frequently, we have long transformations of DataFrames:

```
df = (
    df
    # some transformations 1
    # some transformations 2
    # some transformations 3
    # some transformations 4
)
```

and stepping through them in an interactive testing session is painful;
we have to break up the code like so:

```
df = (
    df
    # some transformations 1
)
breakpoint()
df = (
    df
    # some transformations 2
)
breakpoint()
df = (
    df
    # some transformations 3
)
breakpoint()
df = (
    df
    # some transformations 4
)
```

this can become cumbersome. the proposal here is to register a
`breakpoint()` method on `DataFrame` objects while testing with
`SparklyGlobalContextTest` so you can just sprinkle `.breakpoint()`
method calls within your chain of transformations:

```
df = (
    df
    # some transformations 1
    .breakpoint()
    # some transformations 2
    .breakpoint()
    # some transformations 3
    .breakpoint()
    # some transformations 4
    .breakpoint()
)
```

you can even pass your own function, for example if you don't want to
jump into a debugger but maybe just print some rows:

```
def show_df(df):
    df.show(10, False)

df = (
    df
    # some transformations 1
    .breakpoint(show_df)
    # some transformations 2
    .breakpoint(show_df)
    # some transformations 3
    .breakpoint(show_df)
    # some transformations 4
    .breakpoint(lambda df: df.show(20, False))
)
```

sparkly/testing.py

@@ -32,6 +32,7 @@
 import warnings
 
 from pyspark.context import SparkContext
+from pyspark.sql import DataFrame
 from pyspark.sql import types as T
 
 from sparkly import SparklySession
@@ -97,6 +98,34 @@ def _ensure_gateway_is_down():
     os.environ.pop('PYSPARK_GATEWAY_SECRET', None)
 
 
+def dataframe_breakpoint(self, break_function=None):
+    """Injected method for DataFrame class during testing
+
+    User may supply their own break_function which takes a single parameter, df
+
+    Example:
+        output_df = (
+            input_df
+            .withColumn(...)
+            .where(...)
+            .groupBy(...)
+            .agg(...)
+            .breakpoint()    # will bring up the pdb debugger
+            .select(...)
+            .breakpoint(lambda df: df.show(10, False))  # will print 10 rows to console
+            ...
+        )
+    """
+
+    def pdb_breakpoint(df):
+        import pdb
+        pdb.set_trace()
+
+    break_function = break_function or pdb_breakpoint
+    break_function(self)
+    return self
+
+
 class SparklyTest(TestCase):
     """Base test for spark scrip tests.
 
@@ -158,6 +187,9 @@ def setUpClass(cls):
 
         cls._init_session()
 
+        # define a `df.breakpoint()` for testing:
+        DataFrame.breakpoint = dataframe_breakpoint
+
         for fixture in cls.class_fixtures:
             fixture.setup_data()
 
@@ -178,6 +210,7 @@ def setUpClass(cls):
     @classmethod
     def tearDownClass(cls):
         cls.spark.stop()
+        delattr(DataFrame, 'breakpoint')
         _ensure_gateway_is_down()
         super(SparklyTest, cls).tearDownClass()