Use more data when trying to determine encoding.

32 bytes wasn't resulting in the wrong determination being made in more
situations than acceptable.  Bumped it up to 4K since most filesystem
reads are at least that size.

markdown2ctags.py

@@ -34,7 +34,7 @@ class ScriptError(Exception):
 
 def detect_encoding(filename):
     with open(filename, 'rb') as f:
-        raw = f.read(32)
+        raw = f.read(4096)
 
     potential_bom = raw[:4]
     bom_encodings = [('utf-8-sig', codecs.BOM_UTF8),