test(stellar): stabilize fuzz harness and add PR sample

[boalambo]

Summary

Stabilizes the Stellar property fuzz harness and adds a quick sample for pull request checks.

Changes

✅ Fix scheduled fuzz timeout

• Increased testTimeout from 20 minutes to 60 minutes in vitest.config.ts
• The 100,000-run fuzz campaign (~6.5min execution + reporting overhead) was timing out during result reporting

✅ Add PR sample test

• New test:fuzz:sample script runs 100 iterations with fixed seed 12345
• Fast enough for pull request checks while still exercising all 12 properties
• Deterministic results for consistent CI behavior

✅ Failure replay instructions

• Added onFailure callback to print the failing seed and exact replay command
• Added console output when using fixed seed to show replay instructions
• Makes debugging property test failures straightforward

✅ Update CI workflow

• PRs now run pnpm test:fuzz:sample after the main test suite
• Scheduled nightly job continues to run the full 100k-run random campaign
• Maintains deep fuzzing coverage without slowing down PR feedback

Testing

The changes have been implemented and committed. The CI will verify:

• Normal test suite passes
• Stellar property sample (100 runs, seed 12345) passes on PRs
• Full fuzz campaign (100k runs) passes on nightly schedule

Related Issue

Closes #247

[drips-wave]

@boalambo Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits

[truthixify]

The bounded PR fuzz sample passes all 12 property tests and keeps the large run in nightly CI. Approved.

[truthixify]

Two regressions remain in the harness change. package.json now contains duplicate test:leak keys, so one definition is silently discarded. Also onFailure is not a fast-check Parameters option, so the new callback is ignored and does not print the promised replay command. Remove the duplicate script and use fast-check's supported reporter/run-details API (including seed and counterexample path) with a deliberately failing test that proves the replay output is emitted.

[boalambo]

Two regressions remain in the harness change. package.json now contains duplicate test:leak keys, so one definition is silently discarded. Also onFailure is not a fast-check Parameters option, so the new callback is ignored and does not print the promised replay command. Remove the duplicate script and use fast-check's supported reporter/run-details API (including seed and counterexample path) with a deliberately failing test that proves the replay output is emitted.

Fixed in 2c5ecfe. Removed the duplicate test:leak (my "Accept both" merge mistake), replaced onFailure with fast-check's reporter, and added a gated proof test (FC_PROOF_TEST=1). Sample run passes 12/12, and the proof test fails with exit 1 and prints the seed (42), path (0:0) and replay command. Ready for re-review.

[truthixify]

The revision removes the duplicate script, replaces the unsupported fast-check onFailure option with a valid reporter, and carries seed plus counterexample path into replay. The fixed-seed sample (12 properties), build, and public type tests all pass locally.