L1 Task 1 — Ready vs Waiting, k_yield vs k_sleep

Four threads in app/src/main.c:

t_low (prio 7), t_med (5), t_high (3) — preemptive, log then sleep
t_coop (prio -1) — cooperative, 5 iterations of busy work then k_yield(), starts 3s late so the preemptive baseline is visible first

Observations

Run frequency follows sleep duration, not priority: T_HIGH logs ~3x as often as T_LOW only because it sleeps 1/3 as long.
Priority only breaks ties. The sole tie is at boot, where the threads log in strict priority order. After that each wakeup lands on its own tick against an empty ready queue.
Measured periods are 110/210/310ms, not 100/200/300ms, because k_msleep() is a minimum and rounds up to the next tick.
Once t_coop starts it starves all three permanently. k_yield() only rotates against threads of equal or higher priority, and t_coop is alone at -1, so it is immediately reselected. The others reach Ready when their timers fire but never run.
k_msleep() would move t_coop to Waiting and let them proceed. That is the Ready vs Waiting distinction.
Log lines split into each other in immediate mode: LOG_INF() writes to the UART synchronously and non-atomically, so a higher-priority thread can preempt mid-message. Visible proof of preemption, but it makes the capture messy.