Changes to reflect all modifications discussed in nltk#924

  where split(' ') attempts to read lines from n-gram files.

* Python 3 compatibility changes to TextCat module: max integer
* Multiple changes in both corpus reader and TextCat module to
  make language detection work on unicode instead of bytestrings
* Removed own version of trigrams in TextCat module, now using
  one from nltk.util

  external dependency on sample text files anymore.

Removed _load_all_ngrams

* Language frequency distributions are now loaded on-demand rather

  than all at init time.

Added TextCat module

* Added check so that NLTK can be loaded but only fails at runtime when the
  regex dependency isn't met.

Added corpus reader for n-gram frequencies #845 #884

  distances calculated. This helps determine if no match was found.

  now automatically calls this method in its constructor.
* Fixed small bug in way I access maxint.
* Removed reference to now deleted method iso_to_friendly() in
  guess_language().

* lang_supported() method has been removed.
* ISO 639-3 to/from human-friendly methods removed.
* Reader will now automatically load all n-gram data for all langs.
* Doctest updated to reflect above changes, 'u' prefix removed in strings,
  clearer wording, minor spelling corrections.

Added corpus reader for n-gram frequencies #845

* ngram_freq method now checks for language support before n-gram lookup.
* Added doctest file.

Crubadan branch update

Handle items in collection that don't have packages

E.g. https://github.com/nltk/nltk/issues/882

nkjp reader