Assorted magic literals: downsample factor 3, repeated 1e-300 log floor, p0/j<177 in s_perturb_primitive

[sbryngelson]

Three independent clusters of unexplained magic literals.

1. Hardcoded downsample factor 3 — int((dim + 1)/3) - 1 with no named constant:

• src/simulation/m_start_up.fpp:299-305 (m_ds, n_ds, p_ds, and *_glb_ds).
• src/simulation/m_data_output.fpp:1657-1659 (m_ds/n_ds/p_ds; the down_sample block also spans L662-740).
  These duplicate the same hardcoded /3 in two files; should be a single named constant (e.g. down_sample_factor = 3).

2. Repeated 1e-300 log floor — toolchain/mfc/viz/interactive.py clamps before log10 in ~8 places (L302, 303, 304, 380, 382, 383, 1854, 1856, 1857; plus related 1e-30 scale floors at L152, 1936, 2369). A single module-level constant (e.g. LOG_FLOOR = 1e-300) would dedupe and document intent.

3. s_perturb_primitive physics-coupled magic — src/pre_process/m_assign_variables.fpp:

• L145: p0 = 101325._wp (atmospheric pressure in Pa — undocumented literal).
• L149: if (j < 177) — couples a physics perturbation to a hardcoded x-index 177, i.e. an undocumented mesh-resolution assumption. This silently does the wrong thing on any grid where index 177 isn't the intended physical location.

Why it's friction: all three are "what does this number mean?" hazards. The j < 177 one is the most dangerous: it bakes a grid assumption into physics with no comment.

Important — do NOT change values: items 1 and 3 feed golden-file-gated output. The fix is to name and comment (p0 → a documented standard-pressure constant; 177 → a named/commented index, ideally derived from grid params), not to alter the numbers.

---

_{Filed from a repo-wide code-cleanliness review; verified against master @ 40dde5e.}

Code references

• src/simulation/m_start_up.fpp:299-305 — downsample /3
• src/simulation/m_data_output.fpp:1657-1659 — downsample /3
• toolchain/mfc/viz/interactive.py:302-304 — 1e-300 log floor
• src/pre_process/m_assign_variables.fpp:145-149 — p0=101325, j<177