Backport wrapping Errno::EADDRNOTAVAIL in the net/http adapter to 0.1x

[embs]

Just as 148b99e also does but for the v0.1x releases.

[embs]

@iMacTia this is a follow up on what we've discussed on gitter.

I wonder what would be the better strategy for testing this change. We don't seem to have unit specs for the Net::HTTP adapter in the 0.1x trunk. It looks like the automated tests undergone a comprehensible refactor in 1.x (I can't see the test/ dir there anymore).

I imagine the following two possibilities for testing this (if desirable right now for the 0.1x):

• introducing unit specs for the adapter and making sure the error is wrapped (I guess this would demand less effort than)
• setting up a new env (in CI) so we can reproduce the issue in an integration test

What do you think?

[iMacTia]

@embs sorry for the late reply.
Yes, you're right, we've abandoned MiniTest in favour of RSpec on the road towards v1.0, so you won't be able to cover this using RSpec. However since this change has been done in master already, I feel comfortable having it backported in 0.x as well 👍