Fix type missmatch for set_opts functions

[vshakitskiy]

See #40 (comment)

[karlsson]

Since the reason type returned by Erlang for ssl:setopts is defines as just term() I guess this is the safest way to go. There is the type socket.SocketReason defined in glisten that seems to cover most reasons (einval for instance) but there is no guarantee it covers all there I think.
What you will lose now is how to see the fault that actually occurred.

[rawhat]

I wish there was a more well-defined list of errors we could get here. It kinda sucks to lose the context here, even if it's just being discarded most of the time. I do think this is more accurate, but I was leaving this open-ended to potentially eventually actually expose these. But I think that can be a problem to solve later.