Fix conversion of hoomd box to vmd (#9)

* Fix conversion of hoomd box to vmd

I think the vmd lattice constants (if A, B, and C are indeed lattice constants) should also involve the tilt factors, consistent with HOOMD's definition: https://hoomd-blue.readthedocs.io/en/stable/box.html. (I noticed in some shear simulations that the box looked too small in the gradient direction!)

* Fixed norm of lattice vectors

Realized I did not perform the proper calculation before...

* Ensure orthorhombic boxes read exactly from GSD

gsdplugin.c

@@ -841,16 +841,20 @@ static int read_gsd_timestep(void *mydata, int natoms, molfile_timestep_t *ts)
             }
         else
             {
-            ts->A = box[0]; ts->B = box[1]; ts->C = box[2];
             if (box[3] != 0.0f || box[4] != 0.0f || box[5] != 0.0f)
                 {
-                // need to resolve the tilt factors into angles
+                // define lattice constants in terms of box size and tilt factors
                 const double xy = (double)box[3];
                 const double xz = (double)box[4];
                 const double yz = (double)box[5];
                 const double norm1 = sqrt(1.0 + xy*xy);
                 const double norm2 = sqrt(1.0 + xz*xz + yz*yz);
 
+                ts->A = box[0];
+                ts->B = (float)(norm1*box[1]);
+                ts->C = (float)(norm2*box[2]);
+
+                // need to resolve the tilt factors into angles
                 const double cos_gamma= xy / norm1;
                 const double cos_beta = xz / norm2;
                 const double cos_alpha = (xy*xz + yz)/(norm1 * norm2);
@@ -861,6 +865,7 @@ static int read_gsd_timestep(void *mydata, int natoms, molfile_timestep_t *ts)
                 }
             else // orthorhombic
                 {
+                ts->A = box[0]; ts->B = box[1]; ts->C = box[2];
                 ts->alpha = 90.0f; ts->beta = 90.0f; ts->gamma = 90.0f;
                 }
             }