Bring back ntob and the call that relies on it in CherryPy tests. Fixes

#58.

CHANGES.rst

@@ -1,3 +1,9 @@
+v5.9.1
+======
+
+- #58: Reverted encoding behavior in wsgi module to correct
+  regression in CherryPy tests.
+
 v5.9.0
 ======
 

cheroot/_compat.py

@@ -5,6 +5,12 @@
 import six
 
 if six.PY3:
+    def ntob(n, encoding='ISO-8859-1'):
+        """Return the given native string as a byte string in the given encoding."""
+        assert_native(n)
+        # In Python 3, the native string type is unicode
+        return n.encode(encoding)
+
     def ntou(n, encoding='ISO-8859-1'):
         """Return the given native string as a unicode string with the given encoding."""
         assert_native(n)
@@ -16,6 +22,13 @@ def bton(b, encoding='ISO-8859-1'):
         return b.decode(encoding)
 else:
     # Python 2
+    def ntob(n, encoding='ISO-8859-1'):
+        """Return the given native string as a byte string in the given encoding."""
+        assert_native(n)
+        # In Python 2, the native string type is bytes. Assume it's already
+        # in the given encoding, which for ISO-8859-1 is almost always what
+        # was intended.
+        return n
 
     def ntou(n, encoding='ISO-8859-1'):
         """Return the given native string as a unicode string with the given encoding."""

cheroot/wsgi.py

@@ -32,7 +32,7 @@ def my_crazy_app(environ, start_response):
 
 from . import server
 from .workers import threadpool
-from ._compat import bton
+from ._compat import ntob, bton
 
 
 class Server(server.HTTPServer):
@@ -161,7 +161,7 @@ def start_response(self, status, headers, exc_info=None):
                     'WSGI response header value %r is not of type str.' % v)
             if k.lower() == 'content-length':
                 self.remaining_bytes_out = int(v)
-            out_header = k.encode('iso-8859-1'), v.encode('iso-8859-1')
+            out_header = ntob(k), ntob(v)
             self.req.outheaders.append(out_header)
 
         return self.write