samples/scheduler/metairq_dispatch: Regression after 30916 (sched: timeout: Do not miss slice timeouts)

[erwango]

Describe the bug
Following merge of 6b17f2f, sample samples/scheduler/metairq_dispatch is now failed on 2 platforms: nucleo_wb55rg and nucleo_f091rc.
Sample works correctly on other boards. Likely the common characteristic is they are the slowest in our bench.

To Reproduce
twister

Expected behavior
Sample should run w/o error

Logs and console output

*** Booting Zephyr OS build v2.5.0-rc1-106-g2712d74cdcbf  ***
I: Starting Thread0 at priority -2
I: Starting Thread1 at priority -1
I: Starting Thread2 at priority 0
I: Starting Thread3 at priority 1
ERROR: Queue full, event dropped!
ERROR: Queue full, event dropped!
ERROR: Queue full, event dropped!
ERROR: Queue full, event dropped!

Environment (please complete the following information):

• OS: Linux
• Toolchain Zephyr SDK
• Commit SHA or Version used: From 6b17f2f

[erwango]

I'm seeing the same effect with `tests/kernel/semaphore/semaphore (only on nucleo_f091rc this time)

 : START - test_sem_measure_timeouts

    Assertion failed at ../../src/main.c:990: test_sem_measure_timeouts: ((en)
time missmatch - expected 1, got 1
 FAIL - test_sem_measure_timeouts

[erwango]

@ceolin I'm available for testing fixes if needed (well, starting monday :-))

[nashif]

@andyross FYI

[andyross]

Can you try #31841 ? That fixes an edge case where a pre-existing bug with slice timeouts got exposed by the fix you bisected to. I was seeing that fail with non-standard clock rates on a renode riscv platform and the tickless_context test. But it's not impossible it's the root cause here too.