Right now CollectionIndexer and CollectionFeature are used for fancy indexing of Graphics in a GraphicCollection, such as LineCollection. However I have found that this indexing, since we're using lists and for loops a lot, becomes very slow with thousands of world objects.

# very slow, 16 seconds to change 1,000
 for g in contours[ixs_visible].graphics:
     if not g.visible:
         g.visible = True
        
 for g in contours[ixs_hide].graphics:
     if g.visible:
         g.visible = False

However if we just use a numpy array of graphics instead of a tuple then it is MUCH faster:

# very fast, 10 ms to change 1,000 on an RX 470
for g in contours.graphics[ixs_visible]:
    if not g.visible:
        g.visible = True
    
for g in contours.graphics[ixs_hide]:
    if g.visible:
        g.visible = False

We can set the array as read-only by setting the writeable flag: https://numpy.org/doc/stable/reference/generated/numpy.ndarray.flags.html#numpy.ndarray.flags

Could create a new class that uses numpy array of Graphics to parse feature changes etc., basically have CollectionIndexer use the numpy array of graphics.