fix mpl 3.11

[ahuang11]

Description

Fixes #850

I fixed a crash on matplotlib 3.11 where a PlateCarree or Mercator plot (also the default projection when features=[...] is passed) rendered blank on its own and raised ValueError: figure size must be positive finite not [ 8. nan] inside a layout. For these two projections GeoViews hides the native x/y axis and draws cartopy gridliner labels instead, and on matplotlib 3.11 Axes._update_title_position computes a non-finite title y for such a hidden-axis GeoAxes, which then propagates through Axes.get_tightbbox into the HoloViews layout sizing and makes the figure size NaN. I freeze the title auto-position with axis._autotitlepos = False in the gridlabels path, gated on matplotlib >= 3.11 through a new MPL_GE_3_11_0 flag in geoviews.util, so the title keeps its finite default location and the earlier behavior is restored. Other projections, matplotlib < 3.11, and the bokeh backend were never affected.

How to test:

import numpy as np, cartopy.crs as ccrs
import geopandas as gpd, geoviews as gv, hvplot.pandas  # noqa
from shapely.geometry import Polygon

gv.extension('matplotlib')
lon = np.linspace(-170, 170, 11); lat = np.linspace(-80, 80, 6)
polys, vals = [], []
for i in range(len(lat) - 1):
    for j in range(len(lon) - 1):
        polys.append(Polygon([(lon[j], lat[i]), (lon[j+1], lat[i]),
                              (lon[j+1], lat[i+1]), (lon[j], lat[i+1])]))
        vals.append(i * 2 + j)
gdf = gpd.GeoDataFrame({'value': np.array(vals)}, geometry=polys, crs='EPSG:4326')
plot = lambda **kw: gdf.hvplot.polygons(color='value', cmap='viridis', **kw)

plot(features=['coastline'])
(plot(projection=ccrs.PlateCarree(), title='PlateCarree')
 + plot(projection=ccrs.Robinson(), title='Robinson'))

AI Disclosure

Tool & Model:
Usage:

• I have tested all AI-generated content in my PR.
• I take responsibility for all AI-generated content in my PR.

Checklist

• Tests added and are passing
• Added documentation

[ahuang11]