Address PR #371 AI review R1: REGISTRY contradiction + surface tightening

P3 fixes from CI reviewer round 1:

- REGISTRY.md `Note (Phase 3 by_path ...)`: drop "Placebos" from the
  "remain sample-level summaries" sentence (TWFE diagnostic stays
  sample-level; placebos are now per-path under the new sub-bullet).
  Resolves the in-note contradiction the reviewer flagged.
- ChaisemartinDHaultfoeuilleResults docstring: add the
  `path_placebo_event_study` Attributes block entry alongside
  `path_effects`, documenting the negative-int inner-key convention
  and the inherited cross-path cohort-sharing deviation.
- TestByPathPlacebo.test_attr_is_none_when_placebo_false: use the
  same `_by_path_placebo_data()` fixture for both placebo-off and
  placebo-on branches, so the difference is attributable solely to
  the `placebo` flag (not a fixture swap).
- TestByPathPlacebo.test_path_placebo_point_estimate_within_path_mean:
  replace the finiteness-only check with an explicit recomputation of
  the within-path-mean DID^pl identity from the raw data, asserting
  equality at atol=1e-10 / rtol=1e-10. Pins the estimand identity
  against silent regressions in the per-path IF construction.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: igerber

diff_diff/chaisemartin_dhaultfoeuille_results.py

@@ -374,6 +374,20 @@ class ChaisemartinDHaultfoeuilleResults:
         ``{"n_groups": int, "frequency_rank": int,
         "horizons": {l: {"effect", "se", "t_stat", "p_value",
         "conf_int", "n_obs"}}}`` for ``l = 1..L_max``.
+    path_placebo_event_study : dict, optional
+        Per-path backward-horizon placebos ``DID^{pl}_{path, l}`` for
+        ``l = 1..L_max``, keyed by observed treatment trajectory (tuple
+        of int). Inner dict keys are **negative** ints (``-l`` for lag
+        ``l``) to mirror the ``placebo_event_study`` convention so a
+        unified ``{**path_effects[p]["horizons"],
+        **path_placebo_event_study[p]}`` view is well-formed across
+        forward and backward horizons. Each inner entry holds
+        ``{"effect", "se", "t_stat", "p_value", "conf_int", "n_obs"}``.
+        Populated when ``by_path`` is a positive int AND
+        ``placebo=True`` AND ``L_max >= 1``; ``None`` otherwise.
+        Inherits the cross-path cohort-sharing SE deviation from R
+        documented for ``path_effects``. See REGISTRY.md
+        ``Note (Phase 3 by_path ...)`` → "Per-path placebos".
     honest_did_results : HonestDiDResults, optional
         HonestDiD sensitivity analysis bounds (Rambachan & Roth 2023).
         Populated when ``honest_did=True`` in ``fit()`` or by calling