[lldb] Use custom script instead of lldb.macosx.crashlog in test

I'm not entirely sure this is the problem, but the Windows bot doesn't
seem to like this test. Let's do something similar to
command_import.test which doesn't have that issue.

lldb/test/Shell/ScriptInterpreter/Python/silent_command_script_import.test

@@ -1,8 +1,11 @@
 # REQUIRES: python
 
-# RUN: %lldb --script-language python -o 'command script import lldb.macosx.crashlog' 2>&1 | FileCheck %s
-# RUN: %lldb --script-language python -o 'command script import -s lldb.macosx.crashlog' 2>&1 | FileCheck %s --check-prefix SILENT
-# RUN: %lldb --script-language python -o 'command script import --silent lldb.macosx.crashlog' 2>&1 | FileCheck %s --check-prefix SILENT
+# RUN: rm -rf %t && mkdir -p %t
+# RUN: echo "print('Rene Magritte')" >> %t/foo.py
 
-CHECK: commands have been installed
-SILENT-NOT: commands have been installed
+# RUN: %lldb --script-language python -o 'command script import %t/foo.py' 2>&1 | FileCheck %s
+# RUN: %lldb --script-language python -o 'command script import -s %t/foo.py' 2>&1 | FileCheck %s --check-prefix SILENT
+# RUN: %lldb --script-language python -o 'command script import --silent %t/foo.py' 2>&1 | FileCheck %s --check-prefix SILENT
+
+CHECK: Rene Magritte
+SILENT-NOT: Rene Magritte