Refactor numerics with higher-order functions

[rossberg]

There is a lot of code duplication in the numerics implementation. This PR simplifies that by factoring out much of the logic into higher-order functions. This is just a start, we can probably simplify further.

However, this is the first use of def's as parameters, which the interpreter does not support yet.

@ShinWonho, @f52985, this PR isn't high priority. Just putting it out there so that you can have a look in case you have spare cycles. :)

[f52985]

Now the interpreter backends support the def parameter!

To keep it simple,
instead of maintaining a new environment,
it uses the original env that interpreter was using,
where key is the concatenation of the prefix $ and the parameter name (i.e. "$f_"),
and the value is the name of the original target function definition. (i.e. FnameV("fadd_"))

This is based on the assumption that the normal variables in SpecTec never starts with "$".

[rossberg]

Cool, thanks @f52985!