Update on "[ExecuTorch][Llama] Change runner to enable chunked prefill"

This diff adds code to chunk prompt longer than max_seq_len to enable prefill of larger context

Differential Revision: [D71833061](https://our.internmc.facebook.com/intern/diff/D71833061/)

[ghstack-poisoned]

Author: kimishpatel

examples/models/llava/runner/llava_runner.cpp

@@ -55,7 +55,8 @@ Error LlavaRunner::load() {
   text_prefiller_ = std::make_unique<llm::TextPrefiller>(
       text_decoder_runner_.get(),
       /*use_kv_cache=*/true,
-      /*enable_parallel_prefill=*/true);
+      /*enable_parallel_prefill=*/true,
+      /*max_seq_len=*/128);
 
   // Load the image prefiller
   image_prefiller_ = std::make_unique<LlavaImagePrefiller>(module_.get());