Add mode, which throws error in case of overflow (COBOL had this mode for many decades!)

[safinaskar]

I propose adding another variant of Decimal128 rounding mode. I will call it "prohibited". It will simply throw an error if exact result cannot be represented as Decimal128. The error should be throw in all cases when resulting mathematical value cannot be stored in Decimal128. This includes all kinds of overflow, underflow and rounding.

Why I want this? Well, because when dealing with money, one often needs to get exact answer or no answer at all. I. e. sometimes visible failure is better than inexact answer.

COBOL had this feature for many decades. COBOL allows one to specify so-called "rounding mode". One of possible rounding modes is "rounding mode prohibited", which is similar to this proposal. Are we worse than COBOL? I don't think so. This would be absurd if we in 2024 don't have this old COBOL feature