Rivescript

- New algorithm for handling variable tags (<get>, <set>, <add>, <sub>,
<mult>, <div>, <bot> and <env>) that allows for iterative nesting of
these tags (for example, <set copy=<get orig>> will work now).
- Fix sorting algorithm, so triggers with matching word counts will be
sorted by length descending.
- stream() function can accept a multiline string instead of an array
- Speed optimization by precompiling as many regexps as possible (what was
especially helpful was to precompile substitution and simple trigger
regexps), taking the time-to-reply for the default brain from ~0.19s down
to ~0.04s
- Add support for `! local concat` option to override concatenation mode
(file scoped)
- Fix the regexp used when matching optionals so that the triggers don't match
on inputs where they shouldn't. (RiveScript-JS issue 46)

- Change package name from python-rivescript to simply rivescript.
- Change from the GPLv2 license to the MIT license.
- Add compatibility with Python 3.
- Add Unicode support for RiveScript documents.
- Prefer the .rive extension for RS documents over the old .rs extension.
- Track filenames and line numbers when parsing RiveScript documents.
- Add Perl object handler example.
- Add current_user() method accessible from inside an object macro.
- Add unit tests.
- Add deparse() function that dumps the active memory state of the bot.
- Add write() method that writes the active memory state back to disk as a
.rive file (uses deparse()).
- Bugfix with substitution placeholders.
- Bugfix with the <input> and <reply> tags.

- Small bugfix in \_rot13 that caused crashes under certain circumstances.
- Small bugfix regarding the {weight} tag and atomic triggers.
- Restructure the RiveScript library (move rivescript.py into a package
folder named 'rivescript', separate the interactive mode code into
interactive.py)

- Initial version of rivescript.py