polygon_query returns only rectangles

Version
spatialdata version 0.6.0

Describe the bug
Hello,
I tried to reproduce a code that was working with a previous version of sdata objects (maybe before zarrV3?). The idea is that I want to select a Region Of Interest (ROI) in a xenium experiment using polygon_query. And recently the results of the query select a whole rectangle section with no consideration to the polygon shape.

I tried to reproduce this behavior with the blob dataset, I hope it can help.

Thanks in advance :)

Expected behavior
I expected to select element that are either entirely or at least touching the ROI and have then in the sdata.tables["table'] only cells within this region. While what I obtained was all cells in some sort of big window.

Working example

    ## Imports 
    import matplotlib.pyplot as plt
    from shapely.geometry import Point, Polygon, MultiPoint, MultiPolygon
    import spatialdata as sd
    from spatialdata.datasets import blobs
    import numpy as np

    ## 0.6.0
    print("version :", sd.__version__)


    sdata = blobs()


    ## Show all cells/blobs and the shape I want to query

    fig, ax = plt.subplots(figsize=(6, 6))
    ## Plot the label image
    im = ax.imshow(sdata.labels["blobs_labels"], origin="lower", cmap="tab20")
    ## Plot the query shape
    sdata.shapes["blobs_polygons"].plot(ax=ax, edgecolor="red", facecolor="none", linewidth=2 )
    plt.title("Polygons used for query over all cells")
    plt.show()



    ## Query the sdata
    sdata_query = sd.polygon_query(sdata, MultiPolygon(sdata.shapes["blobs_polygons"].geometry.values), target_coordinate_system="global")



    ## Show the results of the query
    ## We can see here that the query was made on a big rectangle, not the selected polygons
    fig, ax = plt.subplots(figsize=(6, 6))
    ## Plot the label image
    im = ax.imshow(sdata_query.labels["blobs_labels"] , origin="lower", cmap="tab20")
    plt.title("Results of the query")
    plt.show()


    ## Check if all cells present in the plot are also present in the tables
    ## We can see that all the blobs present in the plot are not filtered from the table, even though some of them were clearly out of the polygons we used to quarry 
    print("Unique cells in the plot :", np.unique(sdata_query.labels["blobs_labels"]))
    print("\nCells present in the table :\n\n", sdata_query.tables["table"].obs)
    ```

Start with the polygon_query entry point and reproduce the report using the blobs dataset and the provided MultiPolygon example. Check why the result behaves like a bounding rectangle, then verify that both labels and sdata.tables["table"] contain only cells entirely within or touching the polygon ROI.