allow yargs.parse() and yargs(foo).argv to accept a raw argument string

[bcoe]

@kellyselden made the great suggestion that yargs be able to handle a raw argument string, prior to it being parsed by the OS. Like so:

var a1 = yargs.parse('-x=\'marks "the" spot\' --foo "break \'dance\'"')

One great advantage to this is that it makes yargs easier to embed in unorthodox environments, .e.g., a REPL in the web-browser.

fixes #327

[kellyselden]

After looking at your tests, it looks like this will work quite well. Nice quick work!

[nexdrew]

@bcoe Sorry about this, I started to take a look a couple different times and never finished my review.

For some reason I thought the tests already used a string of args, so thought we could reuse logic there, but that's not the case - they use an array of strings.

So overall this LGTM. Nice work!

[bcoe]

@nexdrew I was thinking that we could actually update the tests to use a string instead, why don't I do that good use of this new feature.