Sorted out mime initialization.

Every mime related function in main __init__.py has a flag check for the
check if initialization has already done. This is nonsense, since it
should be done implicitly early on the converter is starting.

This commit straight the things out, and initialization is done in cli
module.

Also, function guess_type was removed, since it's just a proxy for
mimetypes.guess_type function.

ebook_converter/ebooks/lrf/input.py

@@ -1,9 +1,10 @@
-import textwrap, operator
-from copy import deepcopy, copy
+import copy
+import mimetypes
+import operator
+import textwrap
 
 from lxml import etree
 
-from ebook_converter import guess_type
 from ebook_converter.polyglot.builtins import as_bytes
 
 
@@ -87,7 +88,7 @@ class MediaType(etree.XSLTExtension):
 
     def execute(self, context, self_node, input_node, output_parent):
         name = input_node.get('file', None)
-        typ = guess_type(name)[0]
+        typ = mimetypes.guess_type(name)[0]
         if not typ:
             typ = 'application/octet-stream'
         output_parent.text = typ
@@ -120,7 +121,7 @@ class TextBlock(etree.XSLTExtension):
         self.plot_map = plot_map
 
     def execute(self, context, self_node, input_node, output_parent):
-        input_node = deepcopy(input_node)
+        input_node = copy.deepcopy(input_node)
         div = etree.Element('div')
         self.render_block(input_node, div)
         output_parent.append(div)
@@ -190,7 +191,7 @@ class TextBlock(etree.XSLTExtension):
             for child in children:
                 p.remove(child)
                 if pattrib and child.tag == "Span":
-                    attrib = copy(pattrib)
+                    attrib = copy.copy(pattrib)
                     attrib.update(child.attrib)
                     child.attrib.update(attrib)