Simplify and fix "removeComments()"

* The separate treatment of comments at the documentElement's level is not necessary - they have a parent (as tested in Python 3.5.0 and 2.7.11 and 2.6.6)! It might not have worked before due to a typo - note the "if isinstance(element,...)" and "len(element.data)" which should both refer to "subelement" instead - or a bug in very old versions of Python).
* Fix the iteration over childNodes (i.e. replace "for subelement in element.childNodes:" with ""for subelement in element.childNodes[:]:". We have to create a copy of the list to iterate over, otherwise we'd be iterating over a list as we change it which leads to unpredictable results.

Fixes #25

scour/scour.py

@@ -2587,20 +2587,11 @@ def removeComments(element) :
    """
    global numCommentBytes
 
-   if isinstance(element, xml.dom.minidom.Document):
+   if isinstance(element, xml.dom.minidom.Comment):
-      # must process the document object separately, because its
-      # documentElement's nodes have None as their parentNode
-      for subelement in element.childNodes:
-         if isinstance(element, xml.dom.minidom.Comment):
-            numCommentBytes += len(element.data)
-            element.documentElement.removeChild(subelement)
-         else:
-            removeComments(subelement)
-   elif isinstance(element, xml.dom.minidom.Comment):
       numCommentBytes += len(element.data)
       element.parentNode.removeChild(element)
    else:
-      for subelement in element.childNodes:
+      for subelement in element.childNodes[:]:
          removeComments(subelement)