Test: Always run the TypedPython leak test now that #98 is fixed

Author: b-re-w

ksp-fixtures/app/src/commonMain/kotlin/fixture/app/TypedPythonBridge.kt

@@ -52,8 +52,10 @@ fun tpCheck(value: Double, limit: Double): Double {
  * wrapper unreachable -- so the token's `sys.getrefcount` is only meaningful after a JVM collection
  * and the cleaner have run (`TypedPythonKotlinCallTest.settledCounts`).
  *
- * Deliberately **not** closed here: closing a `PyObject` argument inside an upcall releases its
- * reference twice and crashed the test JVM (`Py_XDECREF` on the cleaner thread) -- binder issue #98.
+ * Not closed here, so the test also covers the cleaner path. Closing it would be equally correct:
+ * `close()` gives back the wrapper's one reference and the cleaner then has nothing left to release
+ * (`closeArgument`, `PyObjectArgumentReleaseTest`). The JVM crash once blamed on that (issue #98) came
+ * from a test wrapping fake addresses as owned objects, not from the binder.
  */
 fun tpTouch(token: PyObject?): Long {
     TpBridgeLedger.touches += 1

ksp-fixtures/app/src/desktopTest/kotlin/fixture/app/TypedPythonKotlinCallTest.kt

@@ -229,14 +229,6 @@ class TypedPythonKotlinCallTest {
      */
     @Test
     fun tenThousandCompiledCallsLeakNoReferenceAndNoKotlinHandle() {
-        // Blocked by python-multiplatform#98: when the cleaner releases the wrappers that 10,000
-        // upcalls leave behind, the binder releases one reference twice and the JVM dies (SIGSEGV in
-        // Py_XDECREF on Cleaner-0) -- on the interpreted control as much as on the compiled loop.
-        // Re-enable with -Dtypedpython.leakTest=true once #98 is fixed.
-        assumeTrue(
-            "blocked by python-multiplatform#98 (binder double release on the cleaner path)",
-            System.getProperty("typedpython.leakTest") == "true",
-        )
         Python3.exec(
             """
             import gc, sys, sysconfig