Missing maximum_nprobes affects performance even when there is no prefilter

The docs state:

    /// Configures the maximum number of partitions to search in the vector index.
    ///
    /// These partitions will only be searched if we have not found `k` results after
    /// searching the minimum number of partitions.  Setting this to None (the default)
    /// will search all partitions if needed.
    ///
    /// This setting only takes effect when a prefilter is in place.  In that case we
    /// can spend more effort to try and find results when the filter is highly selective.
    ///
    /// If there is no prefilter, or the results are not highly selective, this value will
    /// have no effect.

However, I have the following search:

dataset
    .scan()
    .nearest("vector", q.as_ref(), K)
    .unwrap()
    .minimum_nprobes(NPROBES)
    .maximum_nprobes(NPROBES)
    .refine(REFINE_FACTOR)
    .project(&["vector", "_distance"])
    .unwrap()

If I drop the maximum_nprobes(NPROBES) then the search is 5x slower. Given there are NPROBES * 5 total partitions in the index I'm guessing it is searching all partitions. It should only do this if it needs to search the extra partitions to find the K results. Since there is no prefilter it should be finding the K results immediately and should not need the extra search.

Start from the nearest, minimum_nprobes, and maximum_nprobes entry points shown in the search chain and trace how partition limits are applied without a prefilter. Reproduce the difference with and without maximum_nprobes, then add a regression test demonstrating that an unfiltered search does not scan unnecessary extra partitions.