Add ThermoResult object and diffrax throw option

[smsharma]

Summary

Implements two requested features:

1. ThermoResult object (closes #14)

• BackgroundModel now returns a ThermoResult NamedTuple containing all thermo outputs plus T_start/T_end
• AbundanceModel accepts ThermoResult directly with automatic temperature range detection
• Maintains backward compatibility via tuple unpacking (use [:7] slice)

2. diffrax throw option (closes #17)

• Add throw parameter to both BackgroundModel and AbundanceModel
• Default is True (raises exceptions on solver failure)
• Set to False for parameter scans where some combinations may fail
• Passed through to all internal diffeqsolve calls

Usage examples

# New streamlined API
thermo_result = BackgroundModel()(0.)
abundances = AbundanceModel(nuclear_net)(thermo_result)
# T_start/T_end auto-detected from thermo_result

# Parameter scans with throw=False
thermo_model = BackgroundModel(throw=False)
abundance_model = AbundanceModel(nuclear_net, throw=False)
# Failures return partial results instead of raising

# Backward compatible usage
t, a, rho_g, rho_nu, rho_extra, P_extra, Neff = thermo_model(0.)[:7]
abundances = abundance_model(rho_g, rho_nu, rho_extra, P_extra, t_vec=t, a_vec=a)

Test plan

• Existing tests pass (pytest pytest/test_abundances.py)
• Tuple unpacking backward compatibility verified
• Test ThermoResult acceptance by AbundanceModel
• Test throw=False behavior

🤖 Generated with Claude Code

[smsharma]

Splitting into separate PRs for #14 and #17