[Repo Assist] Fix Wilcoxon signed-rank test: incorrect rank multiplication in tie correction

[github-actions]

🤖 This is an automated PR from Repo Assist.

Closes #347

Root Cause

The tieCorrection function in Testing/Wilcoxon.fs was multiplying the correction term by the rank value i when a tie group had size > 2:

// Before (buggy)
let tieCorrection (i,j) = 
    if j = 2.0 then 
        (j**3. - j) / 48.
    else i * ((j**3. - j) / 48.)  // ← rank `i` multiplied incorrectly

The standard Wilcoxon signed-rank tie correction formula is:

$$\text{correction} = \sum_k \frac{t_k^3 - t_k}{48}$$

where $t_k$ is the number of observations in each tied group — the rank value plays no role. The special case for j = 2.0 happened to be correct (because it skipped the multiplication), but any tie group with 3+ equal values produced an inflated correction term, leading to incorrect z-statistics and p-values.

Fix

// After (correct)
let tieCorrection (_,j) = 
    (j**3. - j) / 48.
```

This aligns with [SciPy's implementation](https://github.com/scipy/scipy/blob/17603e519b3fe2cb3a94dcda99475f3100f23828/scipy/stats/_wilcoxon.py#L168) referenced in both the documentation and the issue.

## Impact

Only affects datasets where at least one value appears **3 or more times** in the ranked differences. Datasets with only pairwise ties (j=2) were unaffected.

## Test Status

All existing Wilcoxon tests continue to pass. Two new regression tests added using `differences = [1, 1, 1, 2, -3]` (a tie group of size 3), validated against SciPy:

```
SciPy wilcoxon([1,1,1,2,-3], mode='approx', correction=False) → p ≈ 0.4922
SciPy wilcoxon([1,1,1,2,-3], mode='approx', correction=True)  → p ≈ 0.5827

✅ Build: succeeded
✅ Tests: 8/8 passed (6 existing + 2 new)

Generated by Repo Assist · ◷

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@d88ca0e8ee2b080fcba4490ac5b657c98a0eb26b