Scalar periodic spline eval ignores x_min

[krystophny]

Problem

The scalar spline path in src/interpolate/interpolate.f90 wraps periodic coordinates using modulo(x, period) and ignores x_min, effectively enforcing a periodic origin at 0.0.

Example (1D):

• src/interpolate/interpolate.f90:108 uses xj = modulo(x, spl%h_step*(spl%num_points-1))
• The batch spline path (src/interpolate/batch_interpolate_1d.f90) instead uses modulo(x - spl%x_min, period) + spl%x_min.

The same scalar pattern appears in the 2D/3D scalar evaluation routines and their derivative variants (multiple sites in src/interpolate/interpolate.f90).

Impact

• Periodic scalar spline evaluation is inconsistent with the batch spline evaluation for nonzero x_min.
• Derivative variants in the scalar path are also inconsistent: some do not apply periodic wrapping at all.

Expected

For periodic splines, scalar spline evaluation should wrap relative to the spline grid origin:

xj = modulo(x - x_min, period) + x_min

and do so consistently across value/derivative routines.

Suggested fix

• Update scalar evaluate_splines_{1d,2d,3d}* routines to wrap with x - x_min (per-dimension).
• Add a targeted unit test that constructs a periodic scalar spline with nonzero x_min and checks equivalence of values (and derivatives where applicable) under shifts by integer multiples of the period.

Notes

This should be fixed before using scalar spline routines as a behavioral harness for GPU-ready refactors.

[krystophny]

Duplicate of #203.