Switch runner to Module::LoadMode::Mmap

All other runners in examples/models/ use Mmap. The two LLM runners
(qwen3_5_moe, gemma4_31b) were the outliers using File, likely
inherited rather than intentional. File uses pread() syscalls per
access; Mmap maps the file lazily and benefits from the OS page cache.
For a 21 GB .pte this matters: per-access pread has measurable overhead
and prevents kernel-level read-ahead.

Co-authored-by: Claude <noreply@anthropic.com>

Author: mergennachin

examples/models/gemma4_31b/main.cpp

@@ -155,7 +155,7 @@ int main(int argc, char** argv) {
   auto module = std::make_unique<Module>(
       FLAGS_model_path,
       data_files,
-      Module::LoadMode::File,
+      Module::LoadMode::Mmap,
       /*event_tracer=*/nullptr,
       /*memory_allocator=*/nullptr,
       /*temp_allocator=*/nullptr,