Missing error handling in `set_skip_timer()` can leak timer state

[K-ANOY]

I found a resource-lifetime issue in the Modbus degradation timer setup path.

File: plugins/modbus/modbus_req.c

Function: set_skip_timer

Relevant code:

typedef struct {
    uint8_t  slave_id;
    uint16_t degrade_time;
} degrade_timer_data_t;

void *degrade_timer(void *arg)
{
    degrade_timer_data_t *data = (degrade_timer_data_t *) arg;

    struct timespec t1 = { .tv_sec = data->degrade_time, .tv_nsec = 0 };
    struct timespec t2 = { 0 };
    nanosleep(&t1, &t2);

    skip[data->slave_id] = false;

    free(data);
    return NULL;
}

void set_skip_timer(uint8_t slave_id, uint32_t degrade_time)
{
    degrade_timer_data_t *data =
        (degrade_timer_data_t *) malloc(sizeof(degrade_timer_data_t));
    data->slave_id     = slave_id;
    data->degrade_time = degrade_time;

    failed_cycles[slave_id] = 0;

    pthread_t timer_thread;
    pthread_create(&timer_thread, NULL, degrade_timer, data);

    pthread_detach(timer_thread);
}

data is allocated for the timer thread and is normally freed by
degrade_timer(). However, set_skip_timer() does not check either allocation
or thread creation:

degrade_timer_data_t *data =
    (degrade_timer_data_t *) malloc(sizeof(degrade_timer_data_t));
data->slave_id = slave_id;
...
pthread_create(&timer_thread, NULL, degrade_timer, data);
pthread_detach(timer_thread);

If malloc() fails, the function immediately dereferences data. If
pthread_create() fails, the timer thread never runs, so data is never freed.
The related skip[slave_id] recovery logic also does not run in that case.

Suggested fix: check malloc() before writing through data; check the return
value from pthread_create() and free data if the thread was not created.
Only call pthread_detach() when pthread_create() succeeds.

[hxy7yx]

fixed in #2901