[Repo Assist] feat: implement OLS.Linear.Univariable.fitWithWeighting (weighted simple linear regression)

[github-actions]

🤖 This is an automated PR from Repo Assist.

Closes #205

Problem

OLS.Linear.Univariable had no weighted variant. Using LinearRegressor.fit with Weighting and Method.SimpleLinear threw a failwithf at runtime. The only workaround was Polynomial.fitWithWeighting 1, which is non-obvious and unnecessarily general.

Root cause

The Univariable module simply never had a weighted implementation.

Fix

Add OLS.Linear.Univariable.fitWithWeighting that solves the WLS normal equations analytically in a single pass:

[Σw    Σwx ] [a]   [Σwy ]
[Σwx   Σwx²] [b] = [Σwxy]

Closed-form solution — no matrix allocation required:

let slope     = (sw * swxy - swx * swy) / (sw * swxx - swx * swx)
let intercept = (swy - slope * swx) / sw

Wire LinearRegressor.fit to use it when Weighting is provided and FittingMethod = SimpleLinear.

Usage

open FSharp.Stats.Fitting.LinearRegression

let xData   = vector [|1.; 2.; 3.; 4.; 5.; 6.|]
let yData   = vector [|4.; 7.; 9.; 10.; 11.; 15.|]
let weights = vector [|1.; 1.; 1.; 1.; 1.; 0.1|]   // down-weight last point

// Direct API
let coef = OLS.Linear.Univariable.fitWithWeighting weights xData yData

// OO API (now works — no longer throws)
let coef2 = LinearRegression.fit(xData, yData, Weighting = weights)
```

## Trade-offs

- The analytical formula uses one pass over the data (O(n), zero heap allocations beyond the final `Coefficients` wrapper), compared to `Polynomial.fitWithWeighting 1` which allocates multiple arrays and calls `solveLinearSystem`.
- The two approaches agree to machine precision (verified by test).
- A degenerate case (all weight concentrated at one x value, giving a singular system) raises `ArgumentException` with a clear message.

## Test Status

6 new tests added, 1199/1199 pass:
```
Passed!  - Failed: 0, Passed: 1199, Skipped: 0, Total: 1199

Tests cover: equal-unit-weight = unweighted, exact coefficient recovery, outlier down-weighting, scale invariance, agreement with Polynomial.fitWithWeighting 1, and LinearRegressor.fit dispatch.

Generated by 🌈 Repo Assist, see workflow run. Learn more.

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@7ee2b60744abf71b985bead4599640f165edcd93