Scan the complete cross-section for the geometry table of a lasym run

[CharlesCNorton]

Fixes #18.

The cross-section table scans theta in [0, pi] at the plane v and takes the other half of the contour from the reflected plane with Z -> -Z, which is the stellarator-symmetric reconstruction. For lasym = T the stored surface already holds the complete contour, so it is now scanned in a single pass over ntheta3 points at the plane itself; the symmetric run keeps its two passes over ntheta2.

Checked with input.cth_like_fixed_bdy run with LASYM = F and LASYM = T: the ellipticity, indentation, d-shape and shift columns agree between the two runs and with the build before the change. On input.cth_like_fixed_bdy_asym (VMEC++ test data) the boundary ellipticity changes from 2.03788 to 2.06080 in the v = 0 plane and from 0.71460 to 0.73748 in the v = pi/Nfp plane, since the lower half of the contour is now the stored one instead of a copy of the upper half. The wout file is unchanged bit for bit.