ENH: numerical overlap and balance diagnostics for TreatmentEffect

I'm opening this as a smaller follow-up to #10224, limited to numerical overlap and covariate balance diagnostics for TreatmentEffect.

Currently, users need to inspect propensity scores and calculate balance statistics themselves to assess whether IPW-based estimates are supported by their data.

The scope here is:

• A summary of unclipped propensity scores by treatment group, including quantiles and counts outside the clipping bounds.
• A covariate balance table with means and standardized mean differences before and after weighting, using the same clipped propensity scores as the estimator.

Plots and explicit trimming remain outside this issue. Trimming changes the target population and needs a separate API and estimand discussion under #10224.

I have an implementation of these numerical diagnostics and will open a PR for this narrower scope.