Take bu0 at its own half-grid surface in the asymmetric current harmonics

[CharlesCNorton]

Fixes #23.

bu0 is the covariant B_theta harmonic at surface js, so it divides by shalf(js). bv0 on the next line and bu0 in the stellarator-symmetric branch above already do; only the asymmetric bu0 used shalf(js+1). ORNL-Fusion/LIBSTELL read_wout_mod.f90 carries shalf(js) here.

Checked with input.cth_like_fixed_bdy_asym (VMEC++ test data) built with and without this one index: currvmns in the wout changes by 0.72 of its maximum, and every other variable in the file is unchanged bit for bit. bu0 enters t2, which enters only currvmns, so nothing else can move. A stellarator-symmetric run does not reach this branch.

VMEC++ takes the inner surface here too, at output_quantities.cc:5498.