Problem

ToNumber / TryToNumber currently parse localized number words into long, so decimal word phrases cannot be represented by the existing API.

Examples that should eventually be supported by a decimal-returning API:

• English: "one point two" -> 1.2
• Hindi: "एक दशमलव दो" -> 1.2
• Urdu: "ایک اعشاریہ دو" -> 1.2

Current behavior

The words-to-number stack is integer-shaped:

• WordsToNumberExtension.ToNumber(...) returns long
• TryToNumber(...) writes out long
• IWordsToNumberConverter.Convert(...) returns long
• shared parsers reduce token phrases into integer sums/scales

That means decimal phrases should currently be rejected rather than parsed through ToNumber.

PR #1738 removed Hindi/Urdu decimal marker words from ignoredTokens so unsupported decimal phrases fail instead of being silently converted to the wrong integer.

Proposed direction

Add explicit decimal word parsing as a separate feature/API, for example:

• ToDecimalNumber(...) / TryToDecimalNumber(..., out decimal value)
• or a parallel IWordsToDecimalNumberConverter contract
• locale data for decimal marker tokens, distinct from filler/conjunction ignoredTokens
• tests for digit-by-digit fractional parsing, leading/trailing fractional zeroes, negatives, malformed phrases, and locale-specific decimal marker words

This should not change the return type or semantics of existing ToNumber, because that would be a breaking API change.