[decimal] Add same_quantum() + ROUND_HALF_DOWN + rounding architecture improvement

[forfudan]

Related to Issue #175.

This PR adds three new features:

1. same_quantum(other) on BigDecimal
   Returns True if both operands have the same scale (exponent), matching Python's Decimal.same_quantum().

2. ROUND_HALF_DOWN rounding mode
   Adds the 7th rounding mode: ties round toward zero (opposite of ROUND_HALF_UP).

3. Rounding architecture improvement
   BigUInt.remove_trailing_digits_with_rounding() and round_to_keep_first_n_digits() (Decimal128) now accept a sign: Bool parameter and handle CEILING/FLOOR → UP/DOWN translation internally
   Removed duplicated CEILING/FLOOR translation from 3 call sites. The else fallback in Decimal128's round_to_keep_first_n_digits no longer silently treats unknown modes as HALF_EVEN — it is now a proper debug_assert for unknown modes. Updated all rounding docstrings to list all 7 modes consistently
   New RoundingMode.half_down() constructor and ROUND_HALF_DOWN comptime alias Rounding logic in BigUInt.remove_trailing_digits_with_rounding(): rounds up only when cut-off digit > 5, or exactly 5 with non-zero digits beyond

[Copilot]

Pull request overview

Adds missing Python decimal.Decimal-compat features to decimo, specifically BigDecimal.same_quantum() and a new ROUND_HALF_DOWN rounding mode.

Changes:

• Add BigDecimal.same_quantum(other) (scale/exponent equality check).
• Introduce ROUND_HALF_DOWN / RoundingMode.half_down() and implement its rounding behavior in BigUInt digit-truncation rounding.
• Extend tests and roadmap docs to cover the new API surface.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
tests/bigdecimal/test_data/bigdecimal_rounding.toml	Adds HALF_DOWN rounding test cases (data-driven).
tests/bigdecimal/test_bigdecimal_rounding.mojo	Adds Python parity tests for ROUND_HALF_DOWN.
tests/bigdecimal/test_bigdecimal_methods.mojo	Adds unit tests for BigDecimal.same_quantum().
src/decimo/rounding_mode.mojo	Adds ROUND_HALF_DOWN alias and RoundingMode.half_down().
src/decimo/biguint/biguint.mojo	Implements HALF_DOWN in remove_trailing_digits_with_rounding().
src/decimo/bigdecimal/rounding.mojo	Documents HALF_DOWN in BigDecimal rounding helpers (one docstring formatting issue noted).
src/decimo/bigdecimal/bigdecimal.mojo	Adds same_quantum() method to BigDecimal.
src/decimo/init.mojo	Exports ROUND_HALF_DOWN from the package entrypoint.
docs/plans/api_roadmap.md	Marks same_quantum and ROUND_HALF_DOWN as implemented.

Comments suppressed due to low confidence (1)

src/decimo/bigdecimal/rounding.mojo:132

• The round_to_precision() docstring rounding-mode list was accidentally collapsed into a single line (HALF_UP + HALF_DOWN + HALF_EVEN), which makes the documentation hard to read and looks like a formatting mistake. Please split these back into separate, properly indented lines (and include HALF_DOWN as its own entry) to match the style used in round() above.

            RoundingMode.ROUND_DOWN: Round toward zero.
            RoundingMode.ROUND_UP: Round away from zero.
            RoundingMode.ROUND_HALF_UP: Round half away from zero.            RoundingMode.ROUND_HALF_DOWN: Round half toward zero.            RoundingMode.ROUND_HALF_EVEN: Round half to even (banker's).
            RoundingMode.ROUND_CEILING: Round toward +∞.
            RoundingMode.ROUND_FLOOR: Round toward -∞.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

RoundingMode.half_down() is added here, but at least one rounding implementation (e.g. src/decimo/decimal128/utility.mojo round_to_keep_first_n_digits) uses an else branch that effectively treats any unhandled mode as HALF_EVEN. After this change, passing RoundingMode.half_down() into Decimal128 rounding will produce HALF_EVEN results instead of HALF_DOWN. Please update those call sites to explicitly handle HALF_DOWN (and ideally make HALF_EVEN an explicit elif, with an error for truly unknown modes).

Suggested change

	"""Round toward zero if <= 0.5."""
	return Self(6)
	"""
	Round toward zero if <= 0.5.
	Note:
	Currently, HALF_DOWN is mapped to the same internal value as HALF_EVEN
	to maintain compatibility with existing rounding implementations that do
	not yet explicitly handle HALF_DOWN. This means that, until those
	implementations are updated, ROUND_HALF_DOWN will behave identically to
	ROUND_HALF_EVEN.
	"""
	return Self(2)