Fix RPi4 failing sel4test in Release mode

[JDuchniewicz]

This fixes the seL4/util_libs#171 issue where in order to have sel4test pass in Release mode for RPi4 one had to comment out GPIO setting. This stems from having just one available allocation to our disposal while we are in serial_utspace_record == true phase (when the serial driver is being set up).

Since there are 2 allocations happening - one for GPIO in uart_gpio_configure using MAP_IF_NULL macro (here), and the other one for UART being done in bcm_uart_init (here). We are running out of available allocations specified by NUM_ALLOC_AT_TO_TRACK.

[lsf37]

Nice. Thank you for tracking that down.

After this is merged, we should enable those tests in ci-actions.

[Indanz]

It only costs a little bit of data to have more than the bare minimum and the define is a maximum, not an exact count.

[Indanz]

As for avoiding the same problem in the future: The only way is to add extra margin and initialise UARTs first before other devices, so the error print works.

[Indanz]

If you rename it, you have to update all users too of course.

  /github/workspace/projects/sel4test/apps/sel4test-driver/src/main.c:515:45: error: ‘NUM_ALLOC_AT_TO_TRACK’ undeclared (first use in this function); did you mean ‘MAX_ALLOC_AT_TO_TRACK’?
    515 |     static uspace_alloc_at_args_t args_prev[NUM_ALLOC_AT_TO_TRACK] = {};
        |                                             ^~~~~~~~~~~~~~~~~~~~~
        |                                             MAX_ALLOC_AT_TO_TRACK