Scheduler hangs when scheduling a task through interrupt service routine about sub-iot-stack HOT 1 CLOSED

The issue seems to be caused by the gpio_int_callback instm32_common_gpio.c starting an atomic section (and thus disabling all interrupts).

The complete flow causing the bug would be as follows:

1. drive_output_high_task is scheduled on the timer
2. the hw timer is configured so it will fire after 0.5 * TIMER_TICKS_PER_SEC
3. The scheduler will then schedule __feed_watchdog_task on the timer (which will stay in the timer queue)
4. Timer fires after 0.5 * TIMER_TICKS_PER_SEC
5. drive_output_high_task is scheduled in the scheduler to be executed as the next task
6. the hw timer is configured to fire for the next task of the timer queue, which at this point is the __feed_watchdog_task. Doing this will set cmp_reg_write_pending to true. This value is set to false again by the timer interrupt (thus indicating the hw timer cmp reg is set properly)
7. drive_output_high_task is executed by the scheduler
8. drive_output_high_task will set the gpio pin high
9. The GPIO interrupt fires
10. gpio_int_callback will start an atomic section (disabling all interrupts!) and calls drive_output_low_ISR
11. drive_output_low_ISR will schedule drive_output_low_task on the timer
12. As drive_output_low_ISR would need to be executed before __feed_watchdog_task, the hwtimer is configured
13. as the timer interrupt has not yet fired at this point to indicate the cmp reg was set successfully, the processor will hang on while(cmp_reg_write_pending)
14. As we are in an atomic section, interrupts are disabled and the timer interrupt will never fire

from sub-iot-stack.