issue in the function computing NDCG

Hi,

I just find there is an issue in the following function
```
def ndcg(y_pred, y_true, top):
assert y_pred.shape[0] == y_true.shape[0]
top = min(top, y_pred.shape[0])

first_k_docs = sorted(zip(y_true, y_pred), key=cmp_to_key(doc_comparator))
first_k_docs = np.array(first_k_docs)[:top,0]

top_k_idxs = np.argsort(y_true)[::-1][:top]
top_k_docs = y_true[top_k_idxs]

dcg = cumulative_gain(first_k_docs)
idcg = cumulative_gain(top_k_docs)

return dcg / idcg if idcg > 0 else 1.
```

how can ndcg=1 if idcg == 0? If idcg == 0 you should just ignore that query. This definitely makes the NDCG look higher than it is expected to be.

Best,

Ruocheng Guo

Start with the ndcg(y_pred, y_true, top) function shown in the issue and locate its definition in the benchmark code. Trace how queries with idcg == 0 are included in the aggregate result; done means those queries are ignored rather than contributing an NDCG of 1, with the affected metric behavior verified.

Written by the indexing model from the issue text.