fix(alkalinity): count fluoride as a conservative strong-acid anion (#458)

[rkingsbury]

Summary

Follow-up to #458 / #459. While investigating the residual ~9% discrepancy between
Solution.alkalinity and the alkalinity reported by the phreeqc2026 engine
(PhrqSol.get_alkalinity()) for seawater, the gap decomposed — via an exact charge-balance
identity — into two independent causes:

1. Fluoride (F⁻) was omitted from the conservative strong-acid-anion set. (~3.4 mg/L as CaCO₃ for seawater)
2. Residual charge imbalance of non-electroneutral inputs, which is a definitional difference
   between the conservative (Stumm) alkalinity and PHREEQC's titration alkalinity. (~7 mg/L for the unbalanced seawater preset)

This PR addresses only cause (1), which is unambiguous: F⁻ is a strong-acid anion (like Cl⁻/Br⁻/I⁻)
and must count against the conservative alkalinity. Cause (2) is left as-is by design — subtracting the
charge imbalance would redefine alkalinity for non-electroneutral inputs (and the conservative and
titration definitions already coincide for any charge-balanced solution). The docstring now notes this
explicitly, and the divergence origin is written up in the #458 discussion.

Changes

• Add ("F", -1.0): -1 to the acid_anions map in Solution.alkalinity.
• Document that the conservative alkalinity equals the engine's titration alkalinity only for a
  charge-balanced solution; non-electroneutral inputs additionally differ by the residual charge imbalance.
• New regression test test_alkalinity_fluoride_is_conservative: a charge-balanced NaF solution has
  zero conservative alkalinity (Na⁺ and F⁻ cancel), and adding 1 mmol/L F⁻ lowers alkalinity by ~50 mg/L as CaCO₃.
• Minor: combine a pre-existing nested if in the engine cross-check (ruff SIM102) and add a CHANGELOG entry.

Verification

• Seawater conservative alkalinity drops from ~117.5 to ~114.1 mg/L as CaCO₃ (the fluoride term),
  matching the charge-balanced PHREEQC value.
• Full tests/test_solution.py, tests/test_engine_phreeqc2026.py, and tests/phreeqc/ suites pass (796 tests).

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.77%. Comparing base (845653a) to head (51a1caa).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #467      +/-   ##
==========================================
- Coverage   87.78%   87.77%   -0.01%     
==========================================
  Files          14       14              
  Lines        1989     1988       -1     
  Branches      347      346       -1     
==========================================
- Hits         1746     1745       -1     
  Misses        192      192              
  Partials       51       51              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[rkingsbury]

@jjstickel I'm going to merge this now to address source 1, but let me know your thoughts about the definitional difference:

2. Residual charge imbalance of non-electroneutral inputs, which is a definitional difference
   between the conservative (Stumm) alkalinity and PHREEQC's titration alkalinity. (~7 mg/L for the unbalanced seawater preset)