fix(common): manually handle DynSolValue json serialization

[0xrusowsky]

• closes: bug(cheatcodes): either serializeJsonType or eip712HashTypedData fails when dealing with large uint128 values #11366

Motivation

the reported bug was originated by the fact that alloy's DynSolType coercion of serde_json::Value into U256 and I256, only supports those values which can be expressed as:

• .as_u64()/.as_i64() for numbers within 64-bit range
• .as_str() for string representations

see impl here: https://github.com/alloy-rs/core/blob/main/crates/dyn-abi/src/eip712/coerce.rs#L59-L77

however, since fn serialize_value_as_json() delegates json serialization to serde, large numbers (> u64::MAX) were being serialized .as_float(). This means that they would return None in the different coercion attempts of the fns described above, and end up throwing an error.

Solution

modified JSON serialization to preserve precision for large numbers:

1. fn serialize_value_as_json(): rather than delegating json serialization to serde, check if numbers fit in i64/u64:
   • fit → serialize as serde_json::Number
   • don't fit → serialize as string to preserve exact value
2. fn json_value_to_token(): needed to be adjusted to not break the fn test_json_roundtrip_guessed test. I tried keeping the functionality as close as possible to the original, by only parsing strings as I256/U256 when they exceed i64/u64 capacity
3. fn parse_json_as(): Support string-to-number coercion for uint/int types

Despite i believe that the applied changes are correct + i tried to minimize the impact of the changes + the fact that they don't strictly break functionality, they alter the behavior and could impact some users.

PR Checklist

• Added Tests
• Added Documentation
• Breaking changes: no, but could change behavior in some cases

[smol-ninja]

Thanks for doing it @0xrusowsky. In which release will it be included?

[grandizzy]

this will be 1.4, we can consider adding it to a 1.3.3 bug fix release if we need to