Store raw, un-decoded subprocess stdout/err bytes for advanced users/debugging?

[bitprophet]

As noted in #350 and other encoding tickets, folks sometimes request access to the raw bytes from the subprocess, instead of the decoded-to-Unicode-strings version that Invoke deals with internally (up to our own stdout, if written there).

This is probably not a big thing to implement, would need to be a separate field on Result, but there are implications re: extra memory used (for large amounts of output) so some thought is necessary re: whether this is an always-available thing or an opt-in behavior.

It's also arguably less necessary now that encoding & organization of related code is marginally more sane than it used to be, and also arguably a case for #353 or similar. I.e. - what would these users be doing with their raw bytes that we didn't do by decoding it using their configured encoding? Shouldn't we just make it easier for that to occur automatically instead? etc.