Merge absorption fix and investigate remaining discrepancies

[skyelaird]

Summary:
Fixed critical bug where ADX (CCIR 252 adjustment) term was negative for E-layer modes, reducing loss instead of adding it. This caused daytime E-layer predictions to be 15-57 dB too optimistic.

Root Cause:
ADX calculation: 1.359 + 8.617 * ln(max(vert_freq/foE, adj_de_loss)) When both vert_freq/foE < 1.0 and adj_de_loss < 1.0, the logarithm was negative, making ADX negative (-4.55 dB), which REDUCED total loss instead of adding extra D-layer absorption.

Fix:
Clamp the argument to ln() to be >= 1.0:

xv = max(mode.ref.vert_freq / self._current_profile.e.fo, self._adj_de_loss)
xv = max(1.0, xv)  # Prevent negative logarithm
adx = self._adj_ccir252_a + self._adj_ccir252_b * np.log(xv)

Impact:

• 7.20 MHz @ 11 UTC: SNR error reduced from +38.0 dB to +22.7 dB (15 dB improvement)
• 9.70 MHz @ 11 UTC: SNR error reduced from +59.4 dB to +2.4 dB (57 dB improvement!)
• Nighttime predictions unchanged (still 2-4 dB errors)

Test Results:
After fix with corrected reference values:

• 7.20 MHz nighttime: -3.7 dB error ✓
• 9.70 MHz nighttime: -2.4 dB error ✓
• 7.20 MHz daytime: +22.7 dB error ⚠️ (improved, still needs work)
• 9.70 MHz daytime: +2.4 dB error ✓ (excellent!)

Also Fixed:

• Corrected test_mode_selection.py reference values (9.70 MHz @ 11 UTC should be 2F2/42dB, not 2E/-15dB as previously listed)

Remaining Issues:

• 7.20 MHz E-layer still 22.7 dB too high (likely due to dev_loss stub)
• dev_loss is hardcoded to 0.0 (see ionospheric_profile.py:710)

See debug_adx_term.py and debug_e_layer.py for detailed analysis.

References #validation #e-layer #absorption #fortran-port