Make seed part of config and pass into recheck data

[njlr]

• Adds the type SeedConfig for defining starting seed configuration
• Extends IPropertyConfig with a SeedConfig
• Adds helpers for setting the SeedConfig in a IPropertyConfig
• Uses SeedConfig to set initial RecheckData in reportWith
• Fixes seed in PropertyConfig.defaults
• Adds a snapshot test for determinism

Relates to #450

[moodmosaic]

Thank you for the PR!

The core feature is a good addition, but the default should be RandomSeed to preserve existing behavior. The withSeed API lets users opt into determinism explicitly, which is the right trade-off for a property testing library.

[moodmosaic]

This is great! Though I think more cases are warranted:

• same fixed seed, same results
• same fixed seed in sync and async path give same results
• withRandomSeed differs from fixed-seed behavior
• recheck path preserves configured seed semantics

[moodmosaic]

This changes the default config from “effectively random every run” to always deterministic.

I'd have to thoroughly check but if the shrinking/recheck model depends on initial seed plus shrink path, changing default seed behavior affects:

• reproduced counterexamples
• golden tests
• user workflows that expect “rerun gives new samples”

---

This is a significant semantic change. Every user who calls Property.check, Property.checkBool, or any function that flows through PropertyConfig.defaults will now:

• Always test the same inputs across runs (unless they opt into withRandomSeed)
• Potentially miss bugs that only manifest with certain random seeds
• Get different behavior from what they've been relying on without changing any of their own code

AFAIK, in most property-based testing systems, the default is random seeds with the seed printed on failure for reproducibility.

The deterministic-by-default philosophy might exist, but flipping an established default can be surprising to the users, if not risky.

[moodmosaic]

That's good: it means only one concrete seed per property run, not reevaluating randomness multiple times.