RadSolve::levelFluxFaceToCenter uses wrong neighbor metric indexing and wrong denominator for `idim==1`

Location

Source/radiation/RadSolve.cpp:672, Source/radiation/RadSolve.cpp:678

Problem

1. r_right is always computed from (i+1,j,k) regardless of direction (idim), which is incorrect for y/z fluxes.
2. In the idim == 1 branch, both terms divide by r_left; the second should divide by r_right.

Impact

• Incorrect geometry-weighted flux averaging in non-x directions.
• Potential directional bias / numerical error in radiation updates.

Steps to Reproduce

1. Run a spherical/cylindrical radiation case where y/z fluxes are significant.
2. Compare against a corrected implementation or symmetry expectation.
3. Observe discrepancies in flux-centered quantities.

Expected Behavior

Neighbor metrics should be computed in the active dimension, and each face term should use its corresponding metric.

Actual Behavior

Right-face metric indexing is hardcoded to x-neighbor; y-branch uses r_left for both terms.

Proposed Fix

int in = i, jn = j, kn = k;
if (idim == 0) ++in;
else if (idim == 1) ++jn;
else ++kn;

edge_center_metric(i, j, k, idim, geomdata, r_left, s_left);
edge_center_metric(in, jn, kn, idim, geomdata, r_right, s_right);

if (idim == 0) {
    t(i,j,k,it) = (f(i,j,k) / (r_left + 1.e-50_rt) + f(i+1,j,k) / r_right) * 0.5_rt;
} else if (idim == 1) {
    t(i,j,k,it) = (f(i,j,k) / r_left + f(i,j+1,k) / r_right) * 0.5_rt;
} else {
    t(i,j,k,it) = (f(i,j,k) + f(i,j,k+1)) * 0.5_rt;
}

Prepared by Codex.

Start at Source/radiation/RadSolve.cpp:672 and :678 in RadSolve::levelFluxFaceToCenter. Check the neighbor metric indexing for each idim and the denominators used by the y branch, then compare a spherical or cylindrical radiation case against symmetry expectations. Done means the active-direction metrics and corresponding face denominators are used without directional bias.