i/o: support writing to a parallel remote stream OR exporting a pyda-free script

[ndrewh]

When actually throwing an exploit, you obviously cannot use dynamic instrumentation to check state on remote. Still, it can be useful to look at state locally (in parallel) while sending data to both remote and local -- remote should be the same, up to ASLR and other environment differences (i.e. environ, etc).

For use in throwers (e.g. for A/D competitions), it might be best if we could export a raw pwntools script that did not use pyda -- but I don't see an easy way to do this.