[3.8] atlc2 parity benchmark suite

[RFingAdam]

Acceptance criteria

• 5 atlc2 published example cases in benchmarks/atlc2_parity.py
• Z₀ within 2% of atlc2's published value
• Rs within 5% of atlc2's published value
• L within 2% of atlc2's published value
• Documented in docs/theory/atlc2_parity.md

Status: ⚠️ Skeleton exists; uses analytical-as-truth instead of atlc2 values

Need actual atlc2 reference outputs (run on a Windows VM, or harvest from
docs / forum posts).