Fix presence events on TCP transport

[recmanj]

What does this PR do?

Fixes presence events by removing a client's presence correctly.

What issues does this PR fix or reference?

Fixes: #62826

Previous Behavior

Presence events not fired when salt-minion status changes.

New Behavior

Presence events fired correctly when salt-minion status changes.

Merge requirements satisfied?

• Docs
• Changelog - https://docs.saltproject.io/en/master/topics/development/changelog.html
• Tests written/updated

Commits signed with GPG?

No

[sakateka]

This patch does even more, it fixes a memory leak.
Because of this regression, a large memory leak appears in salt-master.

USER        PID %CPU %MEM    VSZ   RSS TTY  STAT START   TIME COMMAND
salt  94353  0.1 21.0 78579320 4110420 ?               S    Oct05   3:34  \_ /usr/bin/salt-master PubServerChannel._publish_daemon

Leaking objects:

builtins.frame
collections.deque
salt.transport.tcp.Subscriber
salt.ext.tornado.gen.Runner
salt.ext.tornado.concurrent.Future
salt.ext.tornado.iostream.IOStream
salt.utils.msgpack.Unpacker

The leak was discovered immediately after the update to 3005.
I researched a bit and also came to this place in the code.

Thank you @recmanj

[sakateka]

@dwoz PTAL

[dwoz]

I'm good with this change but it would be nice to have a test which verifies the presence callbacks are getting called.

[garethgreenaway]

@sakateka Just a heads up about the request to add an additional test for this change.

[Ch3LL]

bump @recmanj are you willing to add test coverage here?

[recmanj]

@Ch3LL on its way

[Ch3LL]

We're wanting to tag for 3006 this friday. I would really love to try to get this in. Any luck on adding test coverage?

[recmanj]

Will push them tomorrow afternoon, sorry