Add confidence intervals to CorrelationResult via Fisher z-transform

[devcrocod]

Description

CorrelationResult currently contains coefficient and pValue but no confidence interval. Add Fisher z-transformation CI for Pearson and Spearman correlations:

• z = arctanh(r), SE = 1/√(n−3)
• CI: tanh(z ± z_{α/2} × SE)

This is already planned in SPECIFICATION.md and matches scipy's pearsonr API which provides confidence_interval(). All prerequisites exist: NormalDistribution is already imported in Correlation.kt, and the CorrelationResult constructor is @PublishedApi internal, so adding a field preserves source compatibility.

Example usage

val result = pearsonCorrelation(x, y, confidenceLevel = 0.95)
result.coefficient        // 0.87
result.pValue             // 0.001
result.confidenceInterval // ConfidenceInterval(lower = 0.72, upper = 0.94)

Related to #50