Add support for Num.abs to Dec type

[sekerez]

Resolves this issue.

This PR does the following:

• Add abs support to Dec by defining a new method in the Zig struct, and an associated function.
• Sort a few lists in both a match statement and a list of constants (mapping Rust enums to corresponding zig functions).

Question: what QA is acceptable? I saw there were no tests for the negate method or negateC function in dec.zig, am I missing something?

[sekerez]

I feel like this code could better leverage Rust's type system in at least two ways:

1. Since all "lowlevel" operations are bunched up in the same enum, no pattern matching of this type can be explicitly exhaustive, as e.g. Dec deoesn't implement StrConcat. It would be better to have the LowLevel enum instead be something like:

pub enum LowLevel {
    Num(NumOp),
    Str(StrOp),
    ...
}

enum NumOp {
    Add,
    Mul,
    ...
}

...

2. More of a nitpick, but functions like build_dec_unary_op suggest that traits should be used instead - something like a Num trait that Dec, Int etc. implement.

Does that sound fair? Can I, in a separate PR, refactor the typing in lowlevel.rs accordingly?

[bhansconnect]

I think this is a pretty reasonable idea, but I think numerics would need to be split farther if you are trying to reach exhaustiveness. For example, some ops may run on integer but on float types.

[sekerez]

@bhansconnect do you know why the performance regression test is failing? I'm up to date with main - could any code change plausibly have caused this? If it's noise, should I raise that as an issue?

[bhansconnect]

@bhansconnect do you know why the performance regression test is failing? I'm up to date with main - could any code change plausibly have caused this? If it's noise, should I raise that as an issue?

Benchmarks flake on rare occasion. Should be good to just rerun and submit.

[bhansconnect]

Thanks for this work. Assuming tests don't flake with the latest main merged in, I have set this to autosubmit.