Retain additional adapter information (timing, timeout, etc)

Some adapters (e.g. typhoeus) have detailed timing information (total_time, starttransfer_time, appconnect_time, pretransfer_time, connect_time, namelookup_time, etc) that they may have gotten from the underlying C library, so discovering timing information via middleware (e.g. start = Time.now; app.call(env).on_complete do |env| duration = Time.now - start end) has an inherent loss of fidelity.

Also, when operating in parallel, adapters don't/can't/shouldn't raise exceptions on timeouts (since we don't currently have any mechanism for handling on_complete callback exceptions when running in parallel). One hack for this is to check for response.status == 0, but it would be better to just ask response.timed_out? There may be other bugs/errors that lead to status == 0, and some adapters might stream the response and set the status to non-zero upfront but then timeout while waiting for the response to complete. So response.status == 0 can't equal timeout. We ought to preserve this information on the response.

response.timed_out? #=> true/false
response.time       #=> total time in seconds
response.time_ms    #=> total time in ms
response.times      #=> {total: t0, starttransfer: t1, appconnect: t2, ...}
response.times_ms   #=> {total: t0, starttransfer: t1, appconnect: t2, ...}

Stored on the response under the obvious keys: :timed_out and :times_ms.

I already have a PR at typhoeus/typhoeus#476 to update the typhoeus adapter with env[:typhoeus_timed_out], but it would make more sense to have a more standard mechanism that could be adopted by any adapters. If you think this is a reasonable approach (or have a different approach to suggest), I'll put together a PR for this (including updating some of other adapters).