If foo[i] or bar[i_index] are used in TAB_REMOVE / TAB_FIND, they will
use another variable with local scope declared inside the macro.

for (int i = 0; i < 1; i++) {
    TAB_REMOVE( count, array, foo[i] );
}

will expand to:

for (int i = 0; i < 1; i++) {
    int i_index;
    .....
    for (int i = 0; i < count; i++)
        if (array[i] == foo[i])
        {
            index = i;
            break;
        }
    .....
}

And inner scope i is used to index foo, instead of the outer scope i
which we would expect without knowing the content of the macro.

link with libm, and also with libdl which is needed

This makes use of a 3rd party circular buffer licensed under MIT.

TODO: Pause, TimeGet, Flush

Take in account the time elapsed since the last callback

Needs testing on device

TimeGet is left as an exercise for tomorrow

this can happen when moving buffers between our 2 linked lists

They stopped working in version 2.0 and never made much sense.

This will replace the "audio-device" variable that was hard-coded into
audio output plugins.

Instead of busy waiting and retrying, buffer packets and wait for the next
pf_play() call to write them to the sound card.
We now have 2 linked lists:
    - packets not accepted yet by Enqueue()
    - packets accepted by Enqueue(), not yet written to hardware, and
    waiting to be released in PlayedCallback()