NumericSpec and operations on numeric

[mversic]

          I think it's better to remove the validation and mention in the docs that mul and div can cause the numeric spec to change, and that users need to do their own validation as necessary. This impl assumes too much, there are multiple ways to handle it (round to `self`, round to provided `spec`, return none if accuracy goes out of bounds, etc)

What are the intended use cases of these ops?

Originally posted by @nxsaken in #4976 (comment)

[Erigara]

Initail design for Numeric was to provide minimal amount of operations.
If smt fancy is required user should:

1. convert it to the appropriate type (from mantissa and scale methods)
2. do operation on this type
3. convert back to Numeric

[mversic]

Initail design for Numeric was to provide minimal amount of operations.

and what do you think of the approach taken in the linked PR? where explicit spec is required on every operation.

If smt fancy is required user should:
convert it to the appropriate type (from mantissa and scale methods)
do operation on this type
convert back to Numeric

I wasn't aware of this, the problem was that there was no way to convert Numeric into f64

[mversic]

except that checked_add and checked_sub are now odd ones out