fix: populate NUTS samples_info keys under test-mode bypass (#1260)

`PYAUTO_TEST_MODE=2` skips the sampler in `_fit_bypass_test_mode`,
producing a samples_info stub with only `total_iterations`, `time`,
and `log_evidence`. Tutorial and downstream code that reads NUTS
diagnostics (`ess_min`, `num_samples`, `mean_acceptance`,
`n_divergent`, `n_logl_evals`) crashed with `KeyError`.

Add a `_test_mode_samples_info()` hook on `AbstractSearch`, merged
into the bypass `samples_info`. Override in `BlackJAXNUTS` to return
the diagnostic keys with NaN/0 placeholders — bypass mode never
ran the sampler, so honest empties propagate to the tutorial print.

Fixes the `autofit_workspace/scripts/searches/mcmc.py` smoke test.

Co-authored-by: Jammy2211 <JNightingale2211@gmail.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: 3 people

autofit/non_linear/search/abstract_search.py

@@ -858,14 +858,16 @@ def _fit_bypass_test_mode(
         # (grid search log_evidences, subhalo Bayesian model comparison,
         # scrape aggregator assertions) doesn't crash on None. SamplesPDF
         # reads log_evidence from samples_info.
+        samples_info = {
+            "total_iterations": 1,
+            "time": 0.0,
+            "log_evidence": log_likelihood,
+        }
+        samples_info.update(self._test_mode_samples_info())
         samples = SamplesPDF(
             model=model,
             sample_list=sample_list,
-            samples_info={
-                "total_iterations": 1,
-                "time": 0.0,
-                "log_evidence": log_likelihood,
-            },
+            samples_info=samples_info,
         )
 
         samples_summary = samples.summary()
@@ -888,6 +890,19 @@ def _fit_bypass_test_mode(
 
         return result
 
+    def _test_mode_samples_info(self) -> dict:
+        """
+        Sampler-specific keys to merge into ``samples_info`` when the
+        sampler is bypassed via ``PYAUTO_TEST_MODE=2`` or ``=3``.
+
+        Override in subclasses to add the diagnostic keys that the real
+        run would populate (e.g. NUTS ESS, MCMC autocorrelations) so that
+        tutorial scripts and downstream code can access those keys
+        without ``KeyError``. Use NaN/0 placeholders — the bypass did not
+        actually sample.
+        """
+        return {}
+
     @staticmethod
     def _build_fake_samples(model, parameter_vector, log_likelihood):
         """

autofit/non_linear/search/mcmc/blackjax/nuts/search.py

@@ -378,6 +378,20 @@ def output_search_internal(self, search_internal):
         with open(self.backend_filename, "wb") as f:
             pickle.dump(search_internal, f)
 
+    def _test_mode_samples_info(self) -> dict:
+        return {
+            "num_warmup": int(self.num_warmup),
+            "num_samples": 0,
+            "num_chains": int(self.num_chains),
+            "ess_min": float("nan"),
+            "ess_per_param": [],
+            "mean_acceptance": float("nan"),
+            "n_divergent": 0,
+            "n_logl_evals": 0,
+            "total_walkers": int(self.num_chains),
+            "total_steps": 0,
+        }
+
     def samples_info_from(self, search_internal=None):
         search_internal = search_internal if search_internal is not None else self.backend