fuzz: add dedicated number boundary fuzzer

[membphis]

Background

Current fuzz targets (fuzz_parse_eager, fuzz_parse_lazy) generate arbitrary JSON, but number parsing edge cases are underrepresented due to random generation bias.

JSON Test Suite includes specific number edge cases (i_number_huge_exp.json, i_number_real_overflow.json), but systematic fuzzing of number boundaries is missing.

Goal

Add a dedicated fuzz target that generates numbers at parsing boundaries to stress-test the number decoder.

Scope

Number Patterns to Generate

• Exponent boundaries: 1e308, 1e-308, 1e309 (overflow), 1e-324 (underflow)
• Integer boundaries: 9223372036854775807 (i64::MAX), 9223372036854775808 (overflow)
• Leading zeros: 00, 01, -00, 0.0, 00.0
• Decimal precision: 17+ digit mantissas
• Sign variations: -0, +1 (invalid), --1 (invalid)
• Whitespace adjacency: numbers preceded/followed by various whitespace

Validation

• EAGER mode must reject RFC 8259 violations (leading zeros, + prefix)
• Valid numbers must round-trip through get_f64 / get_i64 correctly
• Overflow detection must be consistent between modes

Acceptance Criteria

• New fuzz target fuzz_numbers in fuzz/fuzz_targets/
• Generator produces numbers biased toward boundary conditions
• Target validates both parse success/failure and decoded value correctness
• Corpus seeded with JSON Test Suite i_number_* cases
• Added to fuzz.yml timed fuzzing schedule

Technical Notes

• Use arbitrary crate to generate structured number components (sign, integer, fraction, exponent)
• Consider property: parse(format!("[{}]", n)).get_f64("[0]") round-trips for valid numbers

References

• RFC 8259 §6 number grammar
• JSON Test Suite number cases: tests/vendor/JSONTestSuite/test_parsing/i_number_*

[coderabbitai]

🔗 Related PRs

#78 - test(fuzz): add nesting-depth stack-safety target [merged]
#141 - test: add FFI grammar corpus [merged]
#144 - test: strengthen lazy phase2 fuzz replay [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!