Fixes #35414 - Bump Sidekiq to 6.3.z

As a side effect, gitlab-sidekiq-fetcher should get updated to >= 7.1.0.

Currently latest sidekiq is 6.5.5, but anything above 6.4.0 is rather
verbose about things being deprecated in sidekiq 7.0.0.

Similarly, latest redis is 4.8.0, but anything above 4.5.0 is rather
verbose about things being deprecated in redis 5.0.0. Note, those
deprecation warnings are triggered from inside of
gitlab-sidekiq-fetcher.

Fixes #35414 - Drop sd_notify usage

Sidekiq should be doing this on its own behind the scenes

Refs #35414 - Fix sd_notify for multiple orchestrators

Refs #35414 - Add sd_notify dependency to service

Puma has a soft dependency on sd_notify when systemd is used. In our
service we use this so we should make sure it's pulled in. Prior to
9b0652fed9f86fd429e237ce359422a1dba14a2a the dynflow_sidekiq group
pulled it in, which technically wasn't entirely correct but in practice
it worked. Sidekiq 6 has essentially vendored sd_notify so it was
removed.

Fixes: 9b0652fed9f86fd429e237ce359422a1dba14a2a

Refs #35414 - Log an early message on dynflow startup

When the service is marked as ready, the verification starts. When
puppet-foreman configures the service using journald logging it checks
the journal. However, the application can take a while to really start
while it's already signaled to systemd it has started up. This means
verification fails.

By logging a message before sending ready we're guaranteed that there is
something to verify.

Fixes: 0ab14da142c9539b0f8de34c32d929d2aa5de6d4

Refs #35414 - Expect a different message in journal

In Foreman 3.5 the Dynflow startup procedure was changed. It now signals
it's ready before it's actually started up. This is because it can run
into passive mode, in case there are multiple orchestrators. This
changes the expectation so it's reliably testing it.