Add fractions module: Rational number arithmetic

Beginne mit der Lektüre von CPythons Lib/fractions.py und dem bestehenden _bigint-Feature-Muster des Projekts, einschließlich cmath hinter dem complex-Feature. Überprüfe das pyo3-proptest-Muster, bevor du die gestufte API und das Verhalten definierst. Erledigt ist die Aufgabe, wenn das fractions-Feature die aufgeführten Konstruktions-, Arithmetik-, Konvertierungs-, Formatierungs- und Edge-Case-Tests gegen fractions.Fraction unterstützt.

Summary

Port Python's fractions.Fraction to Rust — arbitrary-precision rational number arithmetic.

In CPython, this is a pure Python implementation (Lib/fractions.py) that internally stores a (numerator: int, denominator: int) pair, GCD-normalized with a positive denominator.

Design

Type representation

pub struct Fraction<I> {
    numerator: I,
    denominator: I,  // always positive, coprime with numerator
}

Follow the existing _bigint feature pattern to support both num-bigint and malachite-bigint. Implement as generic over integer type or via trait bounds.

Feature flag

[features]
fractions = ["_bigint"]  # requires BigInt

Same pattern as cmath behind the complex feature.

Implementation plan

Phase 1: Core struct & construction

• Fraction struct (numerator, denominator)
• GCD normalization (sign normalization included: denominator always positive)
• new(numerator, denominator) — primary constructor
• from_integer(n) — construct from integer
• from_float(f64) — using f64::integer_decode() or as_integer_ratio logic
• from_coprime_ints(n, d) — direct construction from pre-normalized values (internal use)
• String parsing: "3/4", "1.5", "1e-2", etc.

Phase 2: Arithmetic operations

• Add, Sub, Mul, Div (Rust traits)
• Neg, Abs
• Floor division, modulo, divmod
• Pow — integer exponent returns Fraction / non-integer returns f64

Arithmetic optimization: Knuth TAOCP 4.5.1 approach (pre-compute GCD in multiplication to avoid unnecessary blowup)

Phase 3: Comparison & conversion

• Eq, Ord (cross-multiply comparison)
• as_f64() — float conversion
• trunc(), floor(), ceil(), round() — integer conversions
• is_integer() — checks denominator == 1
• as_integer_ratio() — returns (numerator, denominator) tuple

Phase 4: Special methods

• limit_denominator(max_denominator) — best rational approximation via continued fraction algorithm
• Hash: compatible with Python's _hash_algorithm (modular inverse based)
• Display: "{numerator}/{denominator}" or "{numerator}" when integer
• Format: float-style formatting (.2f, e, g, etc.)

Phase 5: Testing

• Use the existing pyo3 proptest pattern from the project
• Verify bit-identical results against fractions.Fraction
• Edge cases: zero, negatives, very large numbers, float precision boundary values

Out of scope

• Python numbers.Rational ABC (replaced by Rust traits)
• Direct conversion with Decimal type (separate module scope)
• Pickle/copy (Python runtime dependent)

References

• CPython Lib/fractions.py
• https://docs.python.org/3/library/fractions.html