fix(lwdid): preserve fitted overall inference

[shawcharles]

Summary

Implements the bounded result-provenance fix in igerber#733.

LWDiD.fit() already computes and stores the canonical overall ATT and its inference. LWDiDResults.aggregate(\"overall\") previously replaced that result with a separate aggregation built from marginal cohort effects and their standard errors. That could change the point estimate and discarded the composite-regression joint inference where it was available.

This change returns an equivalent result object containing the fitted canonical result instead of reconstructing a second estimator.

This does not establish the validity of the non-composite delta-method variance used by fit() for other staggered paths. That path still omits cross-cohort covariance; this PR ensures aggregate() no longer hides its provenance by replacing it with a different calculation. The remaining statistical work is tracked in igerber#735.

Test

• Adds an end-to-end staggered regression test for the RA/classical/never-treated composite path.
• The test asserts exact agreement of ATT, SE, t-statistic, p-value, confidence interval, and df_inference between fit() and aggregate(\"overall\").
• Adds a slow, strict-xfail unit-cluster-bootstrap concordance test for non-composite staggered inference. It documents the covariance-aware target for LWDiD: use covariance-aware inference for staggered overall ATT igerber/diff-diff#735 and must be removed when that implementation passes it.
• tests/test_lwdid.py: 64 passed.
• The covariance test: expected xfail.
• Black check passed for touched files.

This PR targets the source branch of igerber#588 so it can be merged there directly.

[cubic-dev-ai]

No issues found across 2 files

_{Re-trigger cubic}