Add security consideration for serialization of numbers

[msporny]

Text needs to be added to the security consideration section to warn developers that using certain classes of numbers will result in a lack of interoperability. We should point to at least the following specifications:

• https://datatracker.ietf.org/doc/html/rfc7493#section-2.2
• https://www.rfc-editor.org/rfc/rfc8785#name-serialization-of-numbers

Guidance provided should be:

• Express arbitrary precision numbers (decimals, floats), as strings.
• Express integers larger than what a 32-bit architecture can support as strings.
• Fractions expressed as decimals/floats almost inevitably lead to interoperability issues.
• Note that not all architectures follow IEEE 754 precision guidelines (CUDA).
• Note that implementations MAY provide features that help developers avoid these interoperability issues.

[msporny]

I'm wondering if we should have a default "interop mode" for DI processors that refuse to digitally sign anything that is in the "non-interoperable numbers" category unless you turn interop mode off.

UPDATE: We discussed this during the 2025-07-15 Data Integrity call and decided to NOT specify a interop mode or MUST/SHOULD requirements around avoiding interoperability-challenged numbers.