use raw files for gist images

The gist API casts all file content to text without mentioning this, so images in gists are corrupted.

Always use the bytes from the raw URL for `image/*` mime types.

nbviewer/providers/gist/handlers.py

@@ -1,9 +1,5 @@
-#-----------------------------------------------------------------------------
-#  Copyright (C) 2013 The IPython Development Team
-#
-#  Distributed under the terms of the BSD License.  The full license is in
-#  the file COPYING, distributed as part of this software.
-#-----------------------------------------------------------------------------
+# Copyright (c) Jupyter Development Team.
+# Distributed under the terms of the Modified BSD License.
 
 import json
 
@@ -109,7 +105,12 @@ def get(self, user, gist_id, filename=''):
 
         if filename and filename in files:
             file = files[filename]
-            if file['truncated']:
+            if (file['type'] or '').startswith('image/'):
+                app_log.debug("Fetching raw image (%s) %s/%s: %s", file['type'], gist_id, filename, file['raw_url'])
+                response = yield self.fetch(file['raw_url'])
+                # use raw bytes for images:
+                content = response.body
+            elif file['truncated']:
                 app_log.debug("Gist %s/%s truncated, fetching %s", gist_id, filename, file['raw_url'])
                 response = yield self.fetch(file['raw_url'])
                 content = response_text(response, encoding='utf-8')
@@ -128,6 +129,7 @@ def get(self, user, gist_id, filename=''):
                     **PROVIDER_CTX
                 )
             else:
+                self.set_header('Content-Type', file.get('type') or 'text/plain')
                 # cannot redirect because of X-Frame-Content
                 self.finish(content)
                 return