feat(edecimal): extend parse to decimal floating-point and scientific notation

[Ravenwater]

Background

`edecimal::parse` currently accepts only an integer-format input (regex `[+-]*[0123456789]+`). It rejects:

• Decimal point: `"3.14"` -> fails
• Scientific notation: `"1.5e10"` -> fails

This is a serious limitation for a decimal number system whose entire purpose is exact decimal arithmetic.

`operator>>` also doesn't `setstate(std::ios::failbit)` on parse failure.

Discovered during Phase E of #835.

Scope

1. Extend `parse()` to accept:
   • Integer literals (existing behavior): `"42"`, `"-1000"`
   • Decimal literals: `"3.14"`, `"-2.5"`, `"0.001"`
   • Scientific notation: `"1e10"`, `"1.5e-100"`, `"3.14e+200"`
2. The edecimal type stores decimal digits in base-10; the parse should preserve those digits exactly (no intermediate floating-point conversion).
3. Operator>> hygiene:
   • Add `if (!(istr >> txt)) return istr;` guard
   • `setstate(failbit)` on parse failure

Implementation notes

• Reuse `sw::universal::string_parse::scan_decimal_float` (the Phase A foundation) for tokenization: it returns int_part, frac_part, exp10 as string_views and validates the input.
• The decimal point position + exponent collapse to an effective digit-shift in the base-10 representation, which edecimal can handle natively.

Acceptance criteria

• New regression test `static/elastic/decimal/conversion/string_parse.cpp` (or appropriate path)
• Coverage: integers, decimal literals, scientific notation, negative values, malformed input rejection, operator>> failbit
• Verified on gcc + clang

Related

• feat: create a decimal string parsing API for all number systems #835: parent epic
• feat(utility): constexpr string-parse primitives (Phase A of #835) #838: scan_decimal_float (the foundation tokenizer for this work)
• Phase E (PR forthcoming) ships the operator>> hygiene fix; the parse extension is deferred to this issue.