BigDecimal should have a major version bump

[crankycoder]

The v3.2.3 revision allows passing in float with no precision argument into the BigDecimal initializer.

This allows values that were previously invalid to now be considered valid. In cases where BigDecimal is being used in validation systems or serialization - this is an API breaking change.

I assume BigDecimal uses semver as it uses a 3 digit version - the major version should be increased to v4.

See message thread in #314 for some more context

[tompng]

Float is not an invalid value as BigDecimal(1.2, 0) is accepted.
BigDecimal(1.2) #=> can't omit precision for a Float. (ArgumentError) is not a validation error. It's an error of missing second argument.

I would like to keep the major version until we have a large API change.

[crankycoder]

Float is not an invalid value as BigDecimal(1.2, 0)

My argument is not that passing in float values is invalid when you pass in a precision.

My argument is that passing in a float without a precision was considered invalid and is now considered valid.

I think it is OK that this change was made - but if code threw an exception before and no longer does so - I cannot see how that is anything other than an API breaking change.