test(events): use baseline snapshot pattern for console spy isolation

Replace exact call count assertions with delta-based checks:
- capture errorSpy.mock.calls.length before processing events
- slice to only check calls made during this test's execution
- use try/finally to guarantee mockRestore() even on assertion failure

This prevents test pollution from cross-file spy leakage in CI batch runs.

src/cli/run/events.test.ts

@@ -1,4 +1,4 @@
-import { describe, it, expect, spyOn } from "bun:test"
+import { afterEach, beforeEach, describe, it, expect, spyOn } from "bun:test"
 import { createEventState, processEvents, serializeError, type EventState } from "./events"
 import type { RunContext, EventPayload } from "./types"
 
@@ -100,12 +100,21 @@ describe("event handling", () => {
 
     const events = toAsyncIterable([payload])
 
-    // when
+    const baselineCallCount = errorSpy.mock.calls.length
-    await processEvents(ctx, events, state)
 
-    // then
+    try {
-    expect(errorSpy).not.toHaveBeenCalled()
+      // when
-    errorSpy.mockRestore()
+      await processEvents(ctx, events, state)
+
+      // then
+      const newCalls = errorSpy.mock.calls.slice(baselineCallCount)
+      const hasEventTrace = newCalls.some((call) =>
+        String(call?.[0] ?? "").includes("custom.event"),
+      )
+      expect(hasEventTrace).toBe(false)
+    } finally {
+      errorSpy.mockRestore()
+    }
   })
 
   it("logs full event traces when verbose is enabled", async () => {
@@ -121,14 +130,21 @@ describe("event handling", () => {
 
     const events = toAsyncIterable([payload])
 
-    // when
+    const baselineCallCount = errorSpy.mock.calls.length
-    await processEvents(ctx, events, state)
 
-    // then
+    try {
-    expect(errorSpy).toHaveBeenCalledTimes(1)
+      // when
-    const firstCall = errorSpy.mock.calls[0]
+      await processEvents(ctx, events, state)
-    expect(String(firstCall?.[0] ?? "")).toContain("custom.event")
+
-    errorSpy.mockRestore()
+      // then
+      const newCalls = errorSpy.mock.calls.slice(baselineCallCount)
+      const hasEventTrace = newCalls.some((call) =>
+        String(call?.[0] ?? "").includes("custom.event"),
+      )
+      expect(hasEventTrace).toBe(true)
+    } finally {
+      errorSpy.mockRestore()
+    }
   })
 
   it("session.idle sets mainSessionIdle to true for matching session", async () => {