perf(get_amount): add float fast path for common units

[rkingsbury]

Summary

get_amount() built a pint Quantity and ran context-based .to() conversions (plus qty.check() dimensionality probes) on every call. In the O(N) property sweeps that iterate over all components — mass, charge_balance, conductivity, total_dissolved_solids, alkalinity, hardness, ... — this per-solute pint overhead dominated the cost.

This adds a float fast path:

• _amount_fast() computes the common units (mol, mol/L, mol/kg/m, g, kg, g/L, mg/L, fraction, eq, eq/L) with plain float arithmetic and a single Quantity construction, and returns None for anything else so get_amount() falls through to the unchanged reference path (M, count, %, SI-prefixed variants, ...).
• _get_amount_mag() returns the bare float; the hot loops use it to skip building a Quantity per solute.
• dielectric_constant hoists the mole-fraction denominator out of its loop — O(N) instead of O(N²), since get_amount(.., "fraction") re-summed all components on every call.

alkalinity, hardness, and ionic_strength speed up with no code change — they call get_amount/solvent_mass, which now hit the fast path.

Correctness

The fast-path arithmetic is bit-identical to pint, verified by probing pint's exact operation order — including the non-obvious detail that pint evaluates g/L/mg/L as (n/V)*mw, not n*mw/V (they differ by 1 ULP for e.g. Mg²⁺). M is deliberately left on the pint path (its conversion differs from mol/L by 1 ULP).

• 125 (solute, unit) combos: 0 mismatches in both magnitude and units vs. the forced pint path.
• No unit tests changed; public API unchanged — get_amount keeps its signature, return type, units, and values. The exact == alias assertions in test_get_amount (ppm == mg/L, eq/L == M, ...) still hold.
• Full suite green: 1986 tests pass across test_solution, test_utils, the phreeqc engines, and the density/dielectric/debye/osmotic/activity/volume/salt-matching files (12 skipped / 2 xfail are pre-existing and unrelated).

Incidental fix: get_amount(<absent solute>, "fraction") previously raised UndefinedUnitError on the reference path (the except DimensionalityError there does not catch it); it now returns 0 (dimensionless), the sensible mole fraction of an absent solute. No test exercised this.

Benchmark — 33-component solution (bit-identical output)

property	baseline	modified	speedup
charge_balance	9.46 ms	0.34 ms	28x
mass	12.12 ms	0.57 ms	21x
total_dissolved_solids	14.56 ms	0.94 ms	15x
dielectric_constant	5.76 ms	0.38 ms	15x (O(N²)→O(N))
conductivity	11.80 ms	1.84 ms	6.4x
alkalinity	4.83 ms	0.75 ms	6.4x
hardness	3.36 ms	0.66 ms	5.1x
ionic_strength	0.59 ms	0.28 ms	2.1x

Construction of the same solution also dropped ~36% (93 → 60 ms).

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 87.30159% with 8 lines in your changes missing coverage. Please review.
✅ Project coverage is 87.31%. Comparing base (4f14b17) to head (a59935e).
⚠️ Report is 5 commits behind head on main.

Files with missing lines	Patch %	Lines
src/pyEQL/solution.py	87.30%	6 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #454      +/-   ##
==========================================
- Coverage   87.81%   87.31%   -0.51%     
==========================================
  Files          14       14              
  Lines        1945     2002      +57     
  Branches      338      354      +16     
==========================================
+ Hits         1708     1748      +40     
- Misses        188      200      +12     
- Partials       49       54       +5     

☔ 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.