compat: PandasArray not accepted by many pandas functions

[jbrockmendel]

In #24990 there was a brief discussion about backing Blocks with PandasArray objects. In an experimental local branch I'm trying this and finding a handful of rough edges that can be smoothed out. I'll update this list as I find more

• view method and size property
• dtypes.cast.astype_nansafe breaks on PandasArray
• assert PandasDtype('object') == np.object fails
• is_dtype_equal(dtype, dtype) returns False for PandasDtype objects

[jbrockmendel]

On further reflection, I would rather nanops retain the tighter typing. Closing.