Memory leak/growth in iOS build when pre-filling tokens via generator's `AppendTokenSequences` API

Describe the bug
We are noticing a memory leak/growth in the iOS build when we load historical chats via the generator's AppendTokenSequences API.

Note that the same code works on the Android side.

Here's the snippet of wrapper code in C++ we have in common to Android/iOS:

void ONNXLLMExecutor::add_input_to_generator(const std::string& input) {
  auto sequences = OgaSequences::Create();
  _tokenizer->Encode(input.c_str(), *sequences);
  _generator->AppendTokenSequences(*sequences);
}

Expected behavior
No memory leak/growth when tokens are pre-filled from a historical chat

Screenshots
N/A

Desktop (please complete the following information):

• OS: [e.g. iOS]
• Browser [e.g. chrome, safari]
• Version [e.g. 22]

Smartphone (please complete the following information):

• Device: [e.g. iPhone 15]
• OS: [e.g. iOS 18.5, 17.5]
• Version [v0.7.0 & v0.8.2]

Additional context
Possibly related issue https://github.com/microsoft/onnxruntime-genai/issues/1075 which was auto closed.