fix: heatmap aggregation when kdims are in a Pandas multi-index

[maximlt]

The optimizations recently done in HoloViews and hvPlot to avoid unnecessary Pandas copies have exposed this bug reported first on hvPlot's issue tracker holoviz/hvplot#1467.

import xarray as xr
import pandas as pd
import holoviews as hv
hv.extension('bokeh')

da = xr.DataArray([[0.5, 0.2], [-0.1, 0.5]], dims=["A", "B"], name="C")
df = da.to_dataframe()
heatmap = hv.HeatMap(df, ['A', 'B'], ['C'])
hv.render(heatmap)

Error: KeyError: "None of ['A', 'B'] are in the columns".

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.79%. Comparing base (d9861ae) to head (b4e63c0).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #6499   +/-   ##
=======================================
  Coverage   88.79%   88.79%           
=======================================
  Files         323      323           
  Lines       68731    68742   +11     
=======================================
+ Hits        61028    61039   +11     
  Misses       7703     7703           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.