Add agent rehearsal harness, findings, and per-run evidence for split-loop-with-lambda

Six scripted pi sub-agent runs (3 tool regimes × 2 models) rehearsed the
exercise; deterministic gate = solution test + approved file. Key results:
mandated CLI triples cost on the strong model (4:06 vs 0:39, PASS both);
weak model fails only under the mandate (A+Q), passes when free or optional.
Drift documented with code: scorer.updateSelection left inside the extracted
method (A+Q, C+Q). Harness in tools/rehearse; summary in
rehearsal-findings.md; per-run evidence in rehearsal-runs/; cheat sheet in
rehearse-with-agent.md corrected (lambda-typed parameters unsupported).

ideas/split-loop-with-lambda/exercise/rehearsal-runs/A/measure.txt

@@ -0,0 +1,35 @@
+--- Run A ---
+session file: /tmp/rehearsal/runs/A/session/2026-09-27T14-48-26-884Z_01a0e356-ba04-72d0-9e2e-57b0a6b3e25d.jsonl
+model:                z-ai/glm-5.3-flash
+session span:         245.6 s
+assistant messages:   27
+tool calls:           27 total
+  bash                     19
+  read                     4
+  edit                     4
+tokens input:         109575
+tokens output:        12572
+tokens cacheRead:     288000
+tokens cacheWrite:    0
+tokens reasoning:     8571
+tokens TOTAL (sum):   418718
+--- /usr/bin/time -v (wall clock + max RSS) ---
+	Elapsed (wall clock) time (h:mm:ss or m:ss): 4:05.76
+	Maximum resident set size (kbytes): 262040
+	Exit status: 0
+--- pi stdout (last 15 lines) ---
++        for (Instant interestingDate : interestingDates) {
++            var forecastDate = Instant.parse("2023-04-26T14:00:00.000-07:00");
++            long daysForward = ChronoUnit.DAYS.between(forecastDate, interestingDate);
++            Boolean lookupWeather = dayOffsetToWeather.apply(daysForward);
++            expectedWeather.add(lookupWeather);
++        }
++        return expectedWeather;
++    }
+
+--- a/src/test/java/codingdojo/LongRangeForecastsTest.java   (forced call-site update)
+-        var actual = new LongRangeForecasts().longRangeForecast(quarter);
++        var actual = new LongRangeForecasts().longRangeForecast();
+```
+
+The scorer dependency is now injectable: `longRangeForecast()` delegates with a lambda wrapping the `scorer` field, and tests can pass a stub `LongFunction<Boolean>` instead.