Let rating service publish rating changes to RabbitMQ

[cleborys]

Closes #803.

[cleborys]

Running the tests locally, I see tests/integration_tests/test_matchmaker.py fail from time to time, both on this branch and on develop. It's not always the same test that fails and running in isolation they all seem to pass.

Likewise tests/integration_tests/test_server_instance.py::test_multiple_contexts seems to also fail on develop, at least when I run it locally.

[Askaholic]

Maybe you want to add an integration test? I think I added some helpers for getting rabbitmq messages in my broadcast PR.

[codecov]

Codecov Report

Merging #804 (6d87d04) into develop (5d3c1cf) will decrease coverage by 0.05%.
The diff coverage is 76.47%.

Impacted Files	Coverage Δ
server/broadcast_service.py	100.00% <ø> (ø)
server/game_service.py	99.01% <ø> (-0.01%)	⬇️
server/lobbyconnection.py	94.20% <ø> (ø)
server/rating_service/rating_service.py	95.72% <73.33%> (-1.97%)	⬇️
server/gameconnection.py	95.05% <100.00%> (ø)
server/message_queue_service.py	97.84% <100.00%> (+0.02%)	⬆️
server/timing/timer.py	79.36% <0.00%> (+1.58%)	⬆️

[cleborys]

I added an integration test - turns out there are some floating point issues somewhere in the interaction between the database and trueskill that lead to results like this:

E       AssertionError: assert {'new_rating_...00000002, ...} == {'new_rating_...: 2000.0, ...}
E         Omitting 6 identical items, use -vv to show
E         Differing items:
E         {'old_rating_mean': 2000.0000000000002} != {'old_rating_mean': 2000.0}