mrc: deal with indexed images

If the input image has mode 'P' or something else that's pretty funky,
the code would currently error out. Let's just convert any image format
that is not "standard" (for some definition of standard) to RGB.

internetarchivepdf/mrc.py

@@ -398,6 +398,11 @@ def create_mrc_hocr_components(image, hocr_word_data,
 
     yield mask_arr
 
+    if image.mode not in ('L', 'RGB'):
+        # Special modes like mapped ('P') or other modes we just map to RGB for
+        # simplicity sake
+        image = image.convert('RGB')
+
     image_arr = np.array(image)
 
     t = time()