decimal: add a native arbitrary-precision decimal engine - #8699

A new `rustpython-decimal` crate implements the General Decimal Arithmetic
Specification over `malachite-bigint`: construction and the numeric-string
grammar, the full arithmetic set, comparison and total ordering, the digit-wise
and exponent operations, correctly rounded sqrt/exp/ln/log10/power, and the
format-specification mini-language.

The algorithms are ported from CPython's `Lib/_pydecimal.py`, which is the
executable specification for the parts the standard leaves open; the coefficient
is a `BigUint` with a cached digit count instead of a digit string, so what the
Python does with slicing is done here with divisions by powers of ten.

Conditions follow libmpdec rather than `_pydecimal`: an operation never raises,
it accumulates the conditions it hit into a status word that the caller merges
into a context. That is what lets a trapped signal report every condition an
operation produced rather than only the first.

Assisted-by: Claude Code:claude-fable-5-1

…eview

- next_plus/next_minus/next_toward: replace a hand-rolled exact-add
  helper that scaled operands without a precision bound (causing
  gigabyte-scale allocations for extreme Emin) with a direct call to
  the already-bounded arith::add.
- bigops::div_pow10/split_pow10: short-circuit to zero when the shift
  exceeds the coefficient's digit count, instead of materializing a
  huge power of ten (e.g. int(Decimal('1e-999999999999999999'))).
- Decimal.__round__(ndigits): honor the context's rounding mode
  instead of hardcoding ROUND_HALF_EVEN, matching quantize's default.
- setcontext(): clear the copied context's status flags when
  installing one of the module's template contexts, matching
  _pydecimal's fresh Context.copy() semantics for that path.
- parse_ascii_raw: clamp an out-of-i64-range exponent instead of
  rejecting the string outright, so context-based conversion still
  produces Infinity/Overflow the way an in-range huge exponent would.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

- fmt::format: bound the fixed-point padding derived from a value's
  exponent the same way width/precision are already bounded, so
  format(Decimal('1e-999999999999999999'), 'f') raises ValueError
  instead of attempting a ~10^18-byte allocation.
- Decimal.as_integer_ratio: an exponent whose reduced power-of-five
  denominator doesn't fit a u32 (e.g. Decimal('1e-4294967296')) now
  raises MemoryError instead of panicking past an .expect().

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

CPython's SignalDict mixes in collections.abc.MutableMapping, so
getcontext().traps.update(...)/.setdefault(...) work on top of
__getitem__/__setitem__ alone. The native class only implemented the
mapping protocol itself, so update and setdefault raised
AttributeError. Add update, setdefault, pop, popitem and clear,
matching MutableMapping's default behavior (the latter three always
raise ValueError("signal keys cannot be deleted"), since deletion is
never allowed).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>