Minimize floating point errors

[MartinSStewart]

There's this neat new tool https://herbie.uwplse.org that takes floating point math and determines if there's a way to reorder the expression so that it produces less rounding errors. This rule would port that tool into Elm and then run it against all floating point expressions in your Elm code.

Example of things the rule would report:

a = sqrt (x+1) - sqrt x

This floating point expression can be replaced with 1 / (sqrt (x+1) + sqrt x) in order to reduce precision errors.`

When (not) to enable this rule:

If floating point rounding errors don't matter much, you have floating point math structured in a specific way for readability, or you don't like having a rule that takes minutes to run (this rule is probably better fit for CI)

I am looking for:

• Would you use this rule?
• Would you use this tool but don't think it's a good fit for an elm-review rule?