Remove the isstring hack in lxml treewalker.

This fixes html5lib#115 as a result, as we're no longer type-sniffing strings.

html5lib/treewalkers/lxmletree.py

@@ -87,10 +87,6 @@ def __init__(self, fragment_root, obj):
  self.tail = ensure_str(self.obj.tail)
  else:
  self.tail = None
- self.isstring = isinstance(obj, str) or isinstance(obj, bytes)
- # Support for bytes here is Py2
- if self.isstring:
- self.obj = ensure_str(self.obj)
 
  def __getattr__(self, name):
  return getattr(self.obj, name)
@@ -143,7 +139,7 @@ def getNodeDetails(self, node):
  elif isinstance(node, Doctype):
  return _base.DOCTYPE, node.name, node.public_id, node.system_id
 
- elif isinstance(node, FragmentWrapper) and node.isstring:
+ elif isinstance(node, FragmentWrapper) and not hasattr(node, "tag"):
  return _base.TEXT, node.obj
 
  elif node.tag == etree.Comment: