Consider switching to `goblin` for ELF support

[danielocfb]

We may want to consider switching to using the goblin crate for our ELF parsing needs. As per my understanding it is stable and used in many contexts. Doing so could also help with issues such as #64 and #53. The reasons for rolling our own ELF parser were somewhere along the lines of performance, but never backed by data to the best of my knowledge.

[d-e-s-o]

I am not super eager to embark on this switch, as our ELF parser seems sufficiently mature at this point. That being said, I think the switch to goblin could help with two things:

• PE support (which apparently it has out of the box); it's still a question whether that would be useful on its own or whether we'd need logic to work with some debug format on top of that to really provide value-add
• non-native endian handling