[refurb] Implement redundant-log-base (FURB163)

[tjkuson]

Summary

Implement simplify-math-log as redundant-log-base (FURB163).

Auto-fixes

import math

math.log(2, 2)

to

import math

math.log2(2)

Related to #1348.

Test Plan

cargo test

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[zanieb]

Nice! This lgtm except the rule needs to go through preview.

[bluetech]

IMO the term "redundant" is not quite right (except the math.log(x, math.e) case). The reason to prefer math.log2(x) over math.log(x, 2) is floating-point accuracy, as mentioned in the function docs.

[tjkuson]

Yeah, that's fair. I did think about which name to use when porting the rule, but never found one I fully liked (it has to both describe both the math.log(foo, math.e) case and the math.log(foo, 2) case). I settled for redundant log base to mean 'there is a way of writing this logarithmic function without specifying a base' instead of 'just remove the base argument'.

I also tried to make the rule message be explicit with what its recommendation is, to guard against confusion.

FURB163.py:15:1: FURB163 [*] Replace with `math.log2(1)`

Happy for name to changed, though. I'm not a huge fan of it.

Also, for what it's worth, the corresponding refurb rule simplify-math-log cites readability, not performance, as the motivation. I cited both readability and precision when porting the rule in the documentation (as well as efficiency; I found the fixes ran faster).

EDIT: Maybe superfluous-log-base?

EDIT 2: 'Accuracy', not 'precision', is the correct term to use here.

[zanieb]

I think the name is fine. You could use imprecise-log-base if you really wanted but I imagine for most users the floating point accuracy is not going to be critical.

[bluetech]

Got it on the name. So my only nit is: in the rule description, replace "precise" with "accurate"; both math.log and math.log2 have the same FP precision (double), but the latter is more accurate (for base 2).