Unicode: encoding detection is inaccurate

[cfrings]

The solution proposed for issue #25 does not, in fact, accurately detect file format when BOM is absent.

If the 32 first characters are pure ASCII, then 'ascii' encoding is "detected". When other characters are present, an encoding compatible with that character is chosen, but it needs not be the correct one.

For instance, with character é present in the first line, I get encoding 'cp-1252' on Windows, while other characters are present later in the file that can't be represented in that encoding (box drawing characters).

The classical coding specification -*- coding: utf-8 -*- is not considered at all.