Add functions for argument reduction

[eggrobin]

Currently we have

• A reduction to ]-π/2, π/2[,

Principia/numerics/elliptic_integrals.cpp

Lines 116 to 120 in 90904e3

	// Argument reduction: angle = fractional_part + integer_part * π where
	// fractional_part is in [-π/2, π/2].
	void Reduce(Angle const& angle,
	Angle& fractional_part,
	std::int64_t& integer_part);

• Many reductions to [0, 2π[,

Principia/astronomy/orbital_elements_body.hpp

Line 461 in 90904e3

? Mod(ω, 2 * π * Radian)

Principia/astronomy/orbit_ground_track_body.hpp

Line 94 in 90904e3

: Mod(offset_longitude, 2 * π * Radian);

some of them with preconditions,

Principia/physics/kepler_orbit_body.hpp

Lines 166 to 169 in 90904e3

	// Maps [-π, π] to [0, 2π].
	auto const positive_angle = [](Angle const& α) -> Angle {
	return α > 0 * Radian ? α : α + 2 * π * Radian;
	};

one slightly better one seemingly used only in tests,

Principia/numerics/double_precision_body.hpp

Lines 309 to 316 in 90904e3

	inline DoublePrecision<Angle> Mod2π(DoublePrecision<Angle> const& θ) {
	static DoublePrecision<Angle> const two_π = []() {
	return QuickTwoSum(0x1.921FB54442D18p2 * Radian,
	0x1.1A62633145C07p-52 * Radian);
	}();
	auto const θ_over_2π = θ / two_π;
	return θ - two_π * DoublePrecision<double>(static_cast<int>(θ_over_2π.value));
	}

None of those are correctly rounded, and we keep writing more of that kind of thing (most recently in #3761 (comment)), there reducing to ]-π, π[.

We should centralize the angle reduction functions somewhere in numerics (and make them correctly rounded someday, see #1760).

[pleroy]

A better implementation would be part of #1760.