Fix huge files regression (for undetected mime)

Note that binary_bytes is an array of bytes that never never appears in text. Array taken from encoding.c source from 'file' utility (http://www.darwinsys.com/file/).

middleman-core/lib/middleman-core/util.rb

@@ -24,7 +24,14 @@ module Middleman
 
       ext = ".#{ext}" unless ext.to_s[0] == ?.
       mime = ::Rack::Mime.mime_type(ext, nil) 
-      return false unless mime
+      unless mime
+        binary_bytes = [0, 1, 2, 3, 4, 5, 6, 11, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 28, 29, 30, 31]
+        s = File.read(filename, File.stat(filename).blksize) || ''        
+        s.each_byte do |c|
+          return true if binary_bytes.include?(c)          
+        end
+        return false
+      end
       return false if mime.start_with?('text/')
       return false if mime.include?('xml')
       return false if mime.include?('json')