utf-16 not auto-detected when finding file

[Dave Love]

The encoding of a utf-16 file isn't detected correctly when finding
the file.

To reproduce: push 'coding-category-utf-16-le onto
`coding-category-list', and put "foo" into a new buffer.  C-x h and
M-x encode-coding-region mule-utf-16le-with-signature.  Then,
correctly:
  (detect-coding-region 1 (point-max) t)
    => mule-utf-16le-with-signature-unix

Save it to a file as binary and then C-x C-v.  It's detected as
latin-1 (in a latin-1 locale).  It is read correctly if you specify
the coding system for C-x C-v.