[Bug] Add OpenNARS-style Distributor in Bag

[maxeeem]

Describe the bug

Higher priority items in Bag should have higher frequency of being picked with lower priority items still having a chance at being selected but much lower than high priority items. This is related to attention and resource allocation.

To Reproduce

Behavior in OpenNARS seems different and closer to what I expected. On 10 takes, I consistently get 6/4 instead of 0/10 or 1/9 in PyNARS. On 100 it's about 50/50.

// b = new Bag(100, 1000, nar.narParameters)
public void testBagPriority(final Bag<NullItem,CharSequence> b) {

    final NullItem na = new NullItem(0.9f);
    final NullItem nb = new NullItem(0.8f);

    b.putIn(na);
    b.putIn(nb);
    b.putIn(new NullItem(0.21f));
    b.putIn(new NullItem(0.24f));
    b.putIn(new NullItem(0.25f));
    b.putIn(new NullItem(0.26f));
    b.putIn(new NullItem(0.27f));
    b.putIn(new NullItem(0.28f));
    b.putIn(new NullItem(0.23f));
    b.putIn(new NullItem(0.22f));

    int hi = 0;
    int lo = 0;
    for (int i = 0; i < 10; i++) {
        final NullItem item = b.takeOut();
        if (item == na || item == nb) {
            hi += 1;
        } else {
            lo += 1;
        }
        b.putIn(item);
    }
    System.out.println(hi + "\t" + lo);

Additional context

Add any other context about the problem here.

[ccrock4t]

Hi Maxim, can you clarify the numbers about the current behavior vs. expected behavior in each system?

[maxeeem]

Hi Maxim, can you clarify the numbers about the current behavior vs. expected behavior in each system?

Yeah, so this is following our discussion over email earlier. Basically, if we have a bag with 10 items, 2 high priority and 8 low priority, and then we perform take 10 times, currently in PyNARS the high priority items will get picked less than 10% of the time. In OpenNARS the two high priority items get picked consistently about 60% of the time which is what we want.

Also, see my comment on the PR for some neat graphs :)