There is now a post-processing step which escapes all "/" characters which transcript testing treats as a regex escape if there isn't a "\" to esape it.

So it turns out that we need contextlib2 for Python 3.4 and earlier.

In order to make "help" behave more consistently for decorated and undecorated commands, argparse output is temporarily redirected to self.stdout.  So doing "help history" is similar to "help load".

However, when using the "-h" with argparse commands without using the "help" command, the output from argparse isn't redirected to self.stdout.  Fixing this would be rather difficult and would essentially involve creating a pyparsing rule to detect it at the parser level.

Also:
- Reanamed foo and bar subcommand methods to base_foo and base_bar

History changes:
- Unknown commands are no longer saved in the history
- history command now has a -t option to generate a transcript based on commands in the history

Also:
- Moved examples transcripts from examples to examples/transcripts
- Added a new transcript for use with the pirate.py example

Also added a section on Sub-commands to the documentation.

Also bumped version to 0.8.0.

Modified the do_help() method to behave differently for methods which have been decorated with an argparse ArgumentParser.
This is so that help will properly deal with sub-command help.

Suppose you have a base command "base" which has two sub-commands, "foo" and "bar".  Then "help base" will provide very different help text than "help base foo".

Slightly tweaked the two argparse decorators to set an attribute in the decorated function's dictionary so that the do_help method can know which functions have an ArgumentParser and which do not.

Added a "subcommands.py" example for demonstrating how to create and use subcommands based on argparse and the cmd2 @with_argument_parser decorator.