Graph.lag should support DataFrames

[martinfleis]

We have a weird situation in Graph.lag now. It technically supports doing a lag of multiple columns at once, but only if those are passed as a numpy array. DataFrame input does not work.

import geopandas as gpd
from geodatasets import get_path

from libpysal.graph import Graph

df = gpd.read_file(get_path('geoda south'))
w = Graph.build_contiguity(df)

# passing aș an array
w.lag(df[['HR60', 'HR70', 'HR80', 'HR90']].values)

array([[ 4.60723336,  3.28482738,  3.21367741,  1.23493381],
       [ 2.65699616,  9.44736896, 14.19951668,  3.56709144],
       [ 5.48348091,  7.71768729,  7.21931735,  5.69651071],
       ...,
       [56.8560935 , 54.66925574, 50.48798788, 35.41970065],
       [31.04795542, 33.44684482, 35.63513443, 43.98784183],
       [37.2522942 , 26.51949069, 25.9177717 , 11.0978376 ]],
      shape=(1412, 4))

# passing as a DataFrame

w.lag(df[['HR60', 'HR70', 'HR80', 'HR90']])

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
/var/folders/2f/fhks6w_d0k556plcv3rfmshw0000gn/T/ipykernel_70018/4208695360.py in ?()
----> 1 w.lag(df[['HR60', 'HR70', 'HR80', 'HR90']])

~/dev/pysal/libpysal/libpysal/graph/base.py in ?(self, y, categorical, ties)
   2083         >>> contiguity_r = contiguity.transform("r")
   2084         >>> contiguity_r.lag(y)
   2085         array([4.2, 1.5, 3. , 2.6, 4.5, 0. , 3. , 0. , 0. ])
   2086         """
-> 2087         return _lag_spatial(self, y, categorical=categorical, ties=ties)

~/dev/pysal/libpysal/libpysal/graph/_spatial_lag.py in ?(graph, y, categorical, ties)
     83     if isinstance(y, list):
     84         y = np.array(y)
     85 
     86     if (
---> 87         isinstance(y.dtype, pd.CategoricalDtype)
     88         or pd.api.types.is_object_dtype(y.dtype)
     89         or pd.api.types.is_bool_dtype(y.dtype)
     90         or pd.api.types.is_string_dtype(y.dtype)

~/dev/pysal/.pixi/envs/default/lib/python3.11/site-packages/pandas/core/generic.py in ?(self, name)
   6317             and name not in self._accessors
   6318             and self._info_axis._can_hold_identifiers_and_holds_name(name)
   6319         ):
   6320             return self[name]
-> 6321         return object.__getattribute__(self, name)

AttributeError: 'DataFrame' object has no attribute 'dtype'

I think we can make those checks for categorical more robust and allow this.