Optimize vloc_of_g by skipping negligible analytic terms

[A-006]

Summary
This commit optimizes the vloc_of_g function in vl_pw.cpp by introducing a numerical cutoff for the analytic Fourier transform correction and refining dependencies.

Key Changes

1. Performance Optimization:
   • Introduced a threshold check: if (gx2 > 921) { continue; }.
   • Since exp(-gx2 * 0.25) is less than 1e-100 when gx2 > 921, skipping this term avoids unnecessary exponential calculations and potential floating-point underflow issues for high-frequency G-vectors.

Verification

• The optimization reduces computational overhead for large basis sets without impacting the precision of the local potential, as the truncated values are well below the double-precision machine epsilon.