Fix bug with misleading idle time, v2 clients

[ivan-at-ericsson]

pafd was reporting v2 clients as active every ~15 seconds due to check_idle() incorrectly calling client_active(). Additionally, v2 clients were receiving idle timeouts when they should have none.

This fix removes the client_active() call for v2 clients in check_idle() and corrects the idle_limit parameter passed to client_connect().

Fixes: #57

[ivan-at-ericsson]

not sure if pull-requests emails work so bringing this to @m-ronnblom or @hofors attention. Just in case you like it.

[m-ronnblom]

Could you detail how to reproduce the issue where "v2 clients were receiving idle timeouts when they should have none"?

[ivan-at-ericsson]

I think, and I could be so wrong, that after the initial change in check_idle, one or multiple tests were failing.
But I will double check this again.

[m-ronnblom]

The 'paf' tests passes before this patch, as does the 'libpaf' tests.

Is this commit AI-generated? I haven't had time to dig into this in detail, but I wanted to make sure you (and not just Codex or some similar tool) had gone through the implications of setting a None idle_limit in sd.py.

From what I recall, "faking" v2 clients as active when checking for idle clients was a bit of a hack, but it served a purpose and the active() call likely can't just be removed. Long time ago, so I could well be wrong here.

[ivan-at-ericsson]

no, I haven't explained myself well.
First, I blindly did the change in check_idle and run the tests, then one/some of them (I assume) failed. Then, again I think, searched for "idle" and "limit" strings which lead to change in 354.
I did this 2 weeks ago, just by searching for bugs in the entire Ericsson/*, checking for (old) "bug" labels, without running the actual application and just trusting the unit tests, since you have plenty of those here. Kudos for that.
I will double check this again and get back to you.