perf: Don't do fused-multiply-add on scalars

[orlp]

If both the inputs to the multiply are scalars it doesn't make much sense to do the optimization. I also refactored the optimization code a bit to be cleaner, and made the FMA code work on scalar columns directly rather than materializing to series.

I also spotted a pretty bad bug where FMA wasn't a * b + c but rather a + b * c, but elsewhere in the code we assumed it was a * b + c: https://github.com/orlp/polars/blob/no-scalar-fma/crates/polars-plan/src/plans/conversion/ir_to_dsl.rs#L1021. So I've made it consistent with the typical a * b + c.

[codecov]

Codecov Report

❌ Patch coverage is 90.62500% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.63%. Comparing base (62674c4) to head (4cc783d).
⚠️ Report is 11 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/polars-ops/src/series/ops/fused.rs	50.00%	2 Missing ⚠️
crates/polars-plan/src/plans/optimizer/fused.rs	96.42%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #27479      +/-   ##
==========================================
+ Coverage   81.34%   81.63%   +0.29%     
==========================================
  Files        1838     1840       +2     
  Lines      254455   254672     +217     
  Branches     3180     3180              
==========================================
+ Hits       206978   207894     +916     
+ Misses      46653    45954     -699     
  Partials      824      824              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ritchie46]

Could this bug be triggered? I would expect this to be hit. Can we add a test.