Solve nondeterminism in testing of hs_get_responsible_hsdirs

Only sometimes do our tests call the case of hs_get_responsible_hsdirs where it "wraps around":

    /* Getting the length of the list if no member is greater than the key we
     * are looking for so start at the first element. */
    if (idx == smartlist_len(sorted_nodes)) {
      start = idx = 0;
    }

We should make this consistent.

I plan to do this, but dgoulet or asn should steal it if they'd rather. :)