Test the Jacobian orientation with a quantity the lasym rotation preserves

[CharlesCNorton]

Fixes #27.

The orientation test runs on a boundary that has already been rotated in the poloidal angle when lasym is set, and rtest*ztest is not invariant under that rotation. The m=1 determinant is, and it reduces to rtest*ztest for a stellarator-symmetric boundary, where rbsc and zbcc are zero, so the symmetric path keeps the criterion it had.

Checked with input.up_down_asym_current relabelled by theta -> pi - theta, which leaves the surface alone and reverses the poloidal direction. Before this change the rotation leaves ztest at 3.7e-17, the test cannot fire, and the run takes 7 Jacobian resets and stops at the 2001 iteration cap with fsqr = 2.07e-02. After it the run converges in 1544 iterations to fsqr = 8.7e-12, and volume_p, aspect, Rmajor_p, Aminor_p, ctor, buco, jcurv, presf and phipf agree with the run on the boundary as given to machine precision. What remains is the poloidal gauge: the two runs differ by the residual rotation, which shows up in the grid-sampled extrema rmax_surf and zmax_surf and in iotaf at the 1e-3 level.

input.up_down_asym_current and input.cth_like_fixed_bdy_asym are unchanged bit for bit, so the test is inert on a boundary that is already oriented correctly, and a stellarator-symmetric run never enters the branch.