remove references to Property Graph (#181)

Nightly builds are failing like this:

```text
WARNING: [autosummary] failed to import cugraph.experimental.PropertyGraph.
Possible hints:
* AttributeError: module 'cugraph.experimental' has no attribute 'PropertyGraph'
* ImportError: 
* ModuleNotFoundError: No module named 'cugraph.experimental.PropertyGraph'

Extension error (sphinx.ext.autosummary)!
```

([build link](https://github.com/rapidsai/cugraph-docs/actions/runs/20615065337/job/59206156695#step:12:1876))

Looks like that's a result of rapidsai/cugraph#5374

This removes all references to cuGraph Property Graphs.

It also updates the description of `cugraph-pyg`, matching what @alexbarghi-nv has started over in rapidsai/cugraph-gnn#375

Authors:
  - James Lamb (https://github.com/jameslamb)

Approvers:
  - Brad Rees (https://github.com/BradReesWork)

URL: #181

Author: jameslamb

docs/cugraph-docs/source/api_docs/cugraph/index.rst

@@ -8,7 +8,6 @@ cugraph API Reference
 
     structure
     graph_implementation
-    property_graph
     centrality
     community
     components

docs/cugraph-docs/source/api_docs/cugraph/property_graph.rst

@@ -3,6 +3,5 @@ Data Stores
 .. toctree::
    :maxdepth: 3
 
-   property_graph.md
    knowledge_stores.md
    feature_stores.md

docs/cugraph-docs/source/graph_support/datastores.rst

@@ -11,7 +11,7 @@ The cuGraph repository has been refactored to make it more efficient to build, m
 Libraries supporting GNNs are now located in the `cugraph-gnn repository <https://github.com/rapidsai/cugraph-gnn>`_
 
 * `pylibwholegraph <https://github.com/rapidsai/cugraph-gnn/tree/main/python/>`_ - the `Wholegraph <https://docs.rapids.ai/api/cugraph/nightly/wholegraph/>`_ library for client memory management supporting cuGraph-PyG for even greater scalability
-* `cugraph_pyg <https://github.com/rapidsai/cugraph-gnn/blob/main/readme_pages/cugraph_pyg.md>`_ enables the ability to use cugraph Property Graphs with PyTorch Geometric (PyG).
+* `cugraph_pyg <https://github.com/rapidsai/cugraph-gnn/blob/main/readme_pages/cugraph_pyg.md>`_ provides native implementations of Pytorch Geometric's (PyG's) `GraphStore`, `FeatureStore`, and `Loader` interfaces, unlocking powerful GPU-accelerated graph analytics—including neighborhood sampling, centrality metrics, and community detection—directly within PyG workflows.
 
 `RAPIDS nx-cugraph <https://rapids.ai/nx-cugraph/>`_ is now located in the `nx-cugraph repository <https://github.com/rapidsai/nx-cugraph>`_ containing a backend to NetworkX for running supported algorithms with GPU acceleration.