Add mesh.acvd.* accessor for PyVista pipelines

[banesullivan]

Using pyacvd today means dropping out of a PyVista pipeline to instantiate Clustering, subdivide, cluster, and rebuild. PyVista 0.48 added a dataset accessor protocol (the same pattern pandas and xarray use). This wires pyacvd into it so uniform remeshing is a one-liner:

import pyacvd  # registers ``mesh.acvd``
remesh = mesh.acvd.remesh(20000, subdivide=3)

What's exposed on pv.PolyData.acvd

• remesh(n_clusters, *, subdivide=0, weights=None, fast=False, ...): one-shot uniform remesh. Auto-triangulates and never mutates the source mesh.
• clustering(weights=None, *, subdivide=0): returns a configured pyacvd.Clustering for fine-grained control (plotting clusters, mixing fast/uniform, etc.).
• cluster_ids(n_clusters, ...): per-point cluster ids without rebuilding the surface.
• subdivide(nsub): linear subdivision returning a new PolyData.

Backwards compatibility

The classic pyacvd.Clustering API is untouched and still works on every supported PyVista version (>= 0.37). Accessor registration is gated by a hasattr(pv, "register_dataset_accessor") check, so older PyVista installs are unaffected.

The accessor is also wired up via the pyvista.accessors entry point, so mesh.acvd works without an explicit import pyacvd once the package is installed.

Notes

• Modeled after the pyvista-manifold accessor layout.
• README has a new "PyVista accessor (recommended)" section above the classic example.