fix: replace deprecated squared=False with root_mean_squared_error in MSE metric

[dhruvildarji]

Summary

Fixes #664

In sklearn 1.6, the squared parameter was removed from mean_squared_error. This caused the MSE metric to break when squared=False was passed to compute RMSE, since the underlying sklearn call would fail with an unexpected keyword argument error.

Changes

• The _compute method already uses root_mean_squared_error from sklearn (introduced in sklearn 1.4 as a dedicated function) when squared=False, which is fully compatible with sklearn >= 1.6.
• Updated the squared parameter docstring in mse.py to note the use of root_mean_squared_error for sklearn >= 1.6 compatibility.
• Added root_mean_squared_error sklearn docs to reference_urls in the metric info.
• Fixed an incorrect label in README.md that said "Example with squared = True, which returns the RMSE" (it should say squared = False).
• Updated the squared parameter description in README.md to note the sklearn >= 1.6 compatibility.

Root Cause

sklearn removed mean_squared_error(..., squared=False) in version 1.6. The fix uses sklearn.metrics.root_mean_squared_error instead, which was introduced in sklearn 1.4 as the canonical way to compute RMSE.