attach: wait for exit code from ContainerWait

[laurazard]

- What I did

Follow up to #5295

Such as with docker run, if a user CTRL-Cs while attached to a container, we should forward the signal and wait for the exit from ContainerWait, instead of just returning.

(see docker attach docs for a clear explanation of the desired behavior)

- How I did it

Used an uncancellable context for the ContainerWait call.

- How to verify it

Run the test:

TESTDIRS="./e2e/container/..." TESTFLAGS="-test.run=TestAttachInterrupt" make -f docker.Makefile test-e2e-non-experimental

- Description for the changelog

Fix `docker attach` exiting on `SIGINT` instead of forwarding the signal to the container and waiting for it to exit.

- A picture of a cute animal (not mandatory but encouraged)

[laurazard]

Oop I forgot to commit something 😅

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 61.00%. Comparing base (788e996) to head (7b46bfc).
Report is 2 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5297      +/-   ##
==========================================
- Coverage   61.45%   61.00%   -0.46%     
==========================================
  Files         299      296       -3     
  Lines       20856    20847       -9     
==========================================
- Hits        12818    12717     -101     
- Misses       7122     7212      +90     
- Partials      916      918       +2     

[laurazard]

I've spent too long trying to figure out why I'd broken it on the connhelper-ssh tests until I remembered maybe it didn't work before 🙃 It does work, just not while running inside dind, but it already didn't before. I'll try to figure out how to make the tests pass on CI, but for now I'm skipping them.

[laurazard]

@thaJeztah @vvoland PTAL

[vvoland]

LGTM; although I'd love to have some tests that would also verify if the returned exit code came from the right place (CLI vs container process), for example:

1. docker run --name asdf -d alpine sh -c 'trap "exit 33" SIGINT; sleep 5; exit 34'
2. docker attach asdf
3. send SIGINT to the cli process
4. check if the exit code is 33 and not 130 (SIGINT)

[laurazard]

@vvoland Oh! That was my idea with this test, since if the CLI was exiting the exit code would not be 130 but 0, but your suggestion is better since we won't be overloading 130 which could be either the CLI or the container SIGINT'ing. I'll change it.