fix: resolve NaN value return in stats

[Soojal1807]

Fix: NaN handling in "autocorr_penalty" in stats.py
Added validation to prevent autocorr_penalty from returning NaN in edge cases.

Issues fixed:

1.Missing length check — Function didn't verify at least 2 data points existed. Short or identical-value series make the correlation calculation undefined hence returning NaN.

2.Unvalidated NaN propagation — A NaN coefficient would silently flow into downstream metrics like sharpe_ratio etc. No check existed to catch this before it spread.

Fix: Added early returns — if data points < 2, or coefficient is NaN, return neutral default 1.0 instead.