Refactor & Expand Baryon Model Tests + Add Benchmarks

[nikosarcevic]

The current baryonic-effects tests in pyccl/tests/ cover all three explicit baryon models (Schneider15, Baccoemu, vanDaalen19) and partially cover baryon-related nonlinear power-spectrum options (CosmicEmu MTIV, HMCode Mead2020 feedback).
However, the coverage is uneven, some tests are bundled together, and there are no structured performance/consistency benchmarks across models.
@elisachisari and I have discussed this and we propose a full cleanup + expansion of the baryon test suite.

Tasks

Audit existing baryon tests

• Review:
  • test_baryons.py
  • test_barryons_vd19.py
  • test_baccoemu.py
  • test_cosmicemu.py
  • test_nonlin_camb_power.py (HMCode + AGN feedback)
  • test_cosmology.py (extra_parameters / halofit plumbing)
• Identify:
  • duplicated logic
  • missing edge cases
  • inconsistent patterns across models

1### Restructure the test layout

Introduce “one model per test file” pattern:

• test_baryons_base.py (from_name, cosmology plumbing, basically tests calls)
• test_baryons_schneider15.py
• test_baryons_baccoemu.py
• test_baryons_vandaalen19.py
• test_baryons_cosmicemu_mtiv.py
• test_baryons_hmcode_mead2020.py

This improves readabiltiy and makes it easier to expand bayron support later when we add more models

Add a small benchmark suite

Create a benchmarks/baryons/ folder with scripts that:

• generate P(k) ratios (baryons on/off) at multiple redshifts
• evaluate models on a standard k–a grid
• optionally compare performance vs reference literature curves
• serve as manual validation for FKEM + baryons integrations
• This will help detect regressions + inconsisetncies