MPSC channel has data race

[msimberg]

As reported by thread sanitizer (e.g. in channel_mpsc_shift test):

==================
WARNING: ThreadSanitizer: data race (pid=3631340)
  Read of size 4 at 0x7b04000028c0 by thread T1:
    #0 pika::experimental::base_channel_mpsc<int, pika::concurrency::detail::spinlock>::get(int*) const /home/mjs/src/pika/libs/pika/synchronization/include/pika/synchronization/channel_mpsc.hpp:112 (channel_mpsc_shift_test+0x10fbb9)
    #1 int channel_get<int>(pika::experimental::base_channel_mpsc<int, pika::concurrency::detail::spinlock> const&) /home/mjs/src/pika/libs/pika/synchronization/tests/unit/channel_mpsc_shift.cpp:30 (channel_mpsc_shift_test+0xfe596)
...

  Previous write of size 4 at 0x7b04000028c0 by thread T2:
    #0 pika::experimental::base_channel_mpsc<int, pika::concurrency::detail::spinlock>::set(int&&) /home/mjs/src/pika/libs/pika/synchronization/include/pika/synchronization/channel_mpsc.hpp:129 (channel_mpsc_shift_test+0x10f562)
    #1 void channel_set<int>(pika::experimental::base_channel_mpsc<int, pika::concurrency::detail::spinlock>&, int) /home/mjs/src/pika/libs/pika/synchronization/tests/unit/channel_mpsc_shift.cpp:37 (channel_mpsc_shift_test+0xfe518)
...

This will be suppressed in #972 for the moment. It may the cause of failures like these (on CircleCI arm64 CI configuration: https://app.circleci.com/pipelines/github/pika-org/pika/9926/workflows/7fd304f2-12e3-49fd-8dfa-59798de62a03/jobs/95968?invite=true#step-106-96082_145):

/home/circleci/project/pika/libs/pika/synchronization/tests/unit/channel_mpsc_fib.cpp(79): test 'fib == verify_fibonacci(i)' failed in function 'void consume_numbers(int, pika::experimental::channel_mpsc<bool> &, pika::experimental::channel_mpsc<int> &, pika::experimental::channel_mpsc<int> &)': '2 != 34'
Tests failed. 0/0 sanity checks and 1/24 tests failed.