del view.attr[key] is not working quite right

a = ad.AnnData(
    np.ones((5, 5)),
    obs=pd.DataFrame({"key": list("abcde")}, index=[f"cell{i}" for i in range(5)]),
    obsm={"key": np.ones((5, 10))},
    uns={"key": "value"},
)
a

AnnData object with n_obs × n_vars = 5 × 5
    obs: 'key'
    uns: 'key'
    obsm: 'key'

• Should do something for uns, but does nothing

   v = a[:3]
   del v.uns["key"]
   display(v)
  

   View of AnnData object with n_obs × n_vars = 3 × 5
       obs: 'key'
       uns: 'key'
       obsm: 'key'
  

• Should we actualize the view for .obs?

   v = a[:3]
   del v.obs["key"]
   display(v)
  

   View of AnnData object with n_obs × n_vars = 3 × 5
       uns: 'key'
       obsm: 'key'
  

• Should we warn about going from view to actual for obsm?

   v = a[:3]
   del v.obsm["key"]
   display(v)
  

   AnnData object with n_obs × n_vars = 3 × 5
       uns: 'key'
  

Should be relatively straightforward fixes to the views.

Start by reproducing the three view-deletion examples in the issue with AnnData, focusing on uns, obs, and obsm. Trace the existing view behavior from those operations and determine the intended result for each case, including whether obsm deletion warns about actualizing the view. Done means the three examples behave consistently with the resolved expectations and have regression coverage.