Added support for identities to Impls

[varun10p]

Added new clause to impl syntax which is currently defined by #:identities ( ...) where

<ident-expr> ::= [<name> <lhs-expr> <rhs-expr>]
               | #:exact <expr>
               | #:commutes

and the keyword translation

#:exact <expr> ==> [<generated-name> <expr> (prog->spec <expr>)]
#:commutes ==> [<impl>-commutes (<impl> a b) (<impl> b a)]

The new identities are added to the operator-impl stuct as a hash map.

[pavpanchekha]

Use triple-backticks for multiline code blocks

[pavpanchekha]

The code is mostly good, but why the specific choice of #:commutes and #:exact? Did you look at the existing FP-safe rules and find that this was enough? Or something else?

Also I see there's no code actually using these rules, I assume that's intended?

[varun10p]

The code is mostly good, but why the specific choice of #:commutes and #:exact? Did you look at the existing FP-safe rules and find that this was enough? Or something else?

Also I see there's no code actually using these rules, I assume that's intended?

Those key words seemed like good shortcuts for now. Yes the code using these rules is coming in the future.

[bksaiki]

I think this PR was actually not ready for review (but feedback appreciated from @pavpanchekha). I was hoping to discuss our approach with fp-safe rules at our next meeting to make sure we're on the same page.

[bksaiki]

Looks mostly good. Just a few changes requested here, as well as outside GitHub.

[bksaiki]

This is still strange. What is the story here?

[bksaiki]

Should #:commutes be at the define-operator-impl level? Just a thought

[varun10p]

It should be

[bksaiki]

Hmmm a bit ugly, but mergeable

[bksaiki]

Have we tested if expressions of these forms simplify?

[bksaiki]

Just a few questions/changes

[pavpanchekha]

@varun10p I'm going to assume Brett is handling this so I won't review unless you tell me you want one from me too.