feat: add runtime TypeSpec support

[MilesCranmerBot]

Summary

• add a runtime TypeSpec for existing Julia types or simple generated structs
• install the SymbolicRegression value hooks from compact Julia callbacks
• support PySRRegressor(type_spec=...) with typed Julia arrays and GenericOperatorEnum, with the custom loss scalar type configured by TypeSpec.loss_type
• keep custom-type equations and prediction Julia-backed

Example

spec = TypeSpec(
    "RASPVal",
    fields={"data": "Union{Float64, Vector{Float64}}"},
    init_value="() -> RASPVal(0.0)",
    sample_value="rng -> RASPVal(randn(rng))",
    mutate_value="(rng, value, temperature) -> value",
    count_scalar_constants=1,
    can_optimize=False,
    loss_type="Float64",
)

Verification

• actual String fit/search/predict
• actual one-field and two-field struct fit/search/predict
• compact hook installation and arity validation
• parameter-group completeness
• ordinary numeric fit/predict smoke test
• full pre-commit suite

Current scope

The custom-type path supports serial and multithreaded searches, existing Julia types, and automatic conversion for generated structs with one or more fields. It remains unweighted and single-output. Denoising, feature selection, resampling, and custom expression specs are rejected explicitly.

No SymbolicRegression.jl backend change is required.

[codecov]

Codecov Report

❌ Patch coverage is 93.51852% with 14 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
pysr/sr.py	81.33%	14 Missing ⚠️

📢 Thoughts on this report? Let us know!

[MilesCranmer]

@MilesCranmerBot please address the above

[MilesCranmer]

@MilesCranmerBot also we need 100% diff coverage

[MilesCranmerBot]

Addressed in 9601b5a:

• renamed the TypeSpec setup method to instantiate
• replaced the @eval wrappers with direct jl.seval definitions
• removed the serial-only restriction and added a multithreaded TypeSpec regression test
• added coverage for TypeSpec conversions and callback variants

Checks: normal numeric regression test passes; TypeSpec direct-instantiation coverage is 100%.