Mesh geometry import

[Luochenghuang]

Summary

Adds Python support for the new MESH geometric object type from Luochenghuang/libctl#mesh-geometry-import. This enables importing arbitrary triangulated 3D surfaces (e.g., from STL/OBJ files) directly into meep simulations.

Depends on libctl with MESH support (NanoComp/libctl#74).
Closes #138.

Changes

python/geom.py — Mesh class and file parsers:

• Mesh(vertices, triangles, center=None, material=...) — triangulated surface geometry
• Mesh.from_stl(filename, ...) — binary and ASCII STL import with vertex welding
• Mesh.from_obj(filename, ...) — Wavefront OBJ import with quad triangulation
• No external dependencies (parsers are pure Python using struct and re)

python/typemap_utils.cpp — C++ binding:

• pymesh_to_mesh — extracts vertices and triangles from Python, calls make_mesh_with_center
• "Mesh" case added to py_gobj_to_gobj dispatch

Usage

import meep as mp

# From STL file
teapot = mp.Mesh.from_stl("model.stl", scale=0.1,
                           material=mp.Medium(epsilon=12))

# From vertices and triangles directly
cube = mp.Mesh(
    vertices=[(-0.5,-0.5,-0.5), (0.5,-0.5,-0.5), ...],
    triangles=[(0,2,1), (0,3,2), ...],
    material=mp.Medium(epsilon=12),
)

sim = mp.Simulation(cell_size=cell, geometry=[teapot], resolution=10)

Subpixel smoothing, epsilon grid queries, and all other meep features work automatically.

Test plan

5 integration tests in python/tests/test_mesh.py:

Test	What it verifies
test_stl_import	Binary STL parse + vertex welding (8 verts, 12 faces)
test_point_in_mesh	Known points inside/outside a cube mesh
test_volume	Fill fraction from grid sampling matches analytic
test_eps_averaging	Smoothing produces intermediate epsilon at interfaces
test_mesh_vs_block	Cube mesh epsilon grid matches native Block exactly