Fix compatability with aiohttp>=3

[atx]

Note that this might be a breaking change - not sure what your policy is on that.

[decaz]

@atx can you make some fixes so tests will pass?

[codecov]

Codecov Report

Merging #713 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #713   +/-   ##
=======================================
  Coverage   89.02%   89.02%           
=======================================
  Files          61       61           
  Lines        4646     4646           
  Branches      818      818           
=======================================
  Hits         4136     4136           
  Misses        332      332           
  Partials      178      178

[atx]

Oops, completely forgot about this - fixed.

The remaning test failures seem to be caused by Tornado 5.0 (testing with 4.5.3 works).

[vdboor]

@atx is there any advantage of specifying the timeout over what I've done in PR #735?

At least PR #735 would remain backwards compatible.

[atx]

I don't think there is. aiohttp itself does some more complicated stuff with its timeout code but it's not obvious to me if it really matters.

[mvantellingen]

I think this was a duplicate with #735 right? I've merged that one before reviewing this one (sorry)

[dirkmb]

besides the fact that #735 missed to change the get-method of AsyncTransport, which still uses aiohttp.Timeout
#764 resolves this problem