catch C++ exceptions at FFI boundary, rename utility modules

[mcharytoniuk]

All C++ wrapper functions now catch exceptions and forward error messages to Rust via out_error parameter with LLAMA_LOG_ERROR at each catch point. sample() returns Result<LlamaToken, SampleError> with the original C++ error message preserved. Renamed llama_utility_* modules to descriptive names.

[Copilot]

Pull request overview

This PR hardens the Rust ↔ C++ FFI boundary by converting C++ exceptions into Rust errors (via out-error pointers) and updates token sampling to return a richer SampleResult so grammar completion can be signaled without crashing/throwing. It also renames several “utility” modules to clearer names and updates tests accordingly.

Changes:

• Add exception-catching + error-message forwarding in C++ wrappers and corresponding Rust error-reading helpers.
• Change LlamaSampler::sample() to return Result<SampleResult, SampleError> and propagate grammar-completion as a first-class outcome.
• Rename / reorganize utility modules (e.g. *_utility_* → focused modules like ggml_time_us, mmap_supported, etc.).

Reviewed changes

Copilot reviewed 16 out of 23 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
llama-cpp-bindings/src/sampling.rs	sample() now returns Result<SampleResult, SampleError> and reads C++ error strings from FFI.
llama-cpp-bindings/src/error.rs	Adds SampleResult / SampleError, improves FFI-related error variants/messages.
llama-cpp-bindings-sys/wrapper_common.{h,cpp}	Updates FFI signatures to include out_error, adds sampler_sample, thinking_end_tag, and catches C++ exceptions.
llama-cpp-bindings/src/model.rs	Adds thinking_end_tag() and new tests around sampling + grammar behavior.
llama-cpp-bindings/tests/*.rs	Updates integration tests to handle SampleResult and stop on GrammarCompleted.
llama-cpp-bindings/src/lib.rs	Renames/exports modules and introduces ffi_error_reader + ffi_status_*.
llama-cpp-bindings/src/json_schema_to_grammar.rs	Uses new out-error plumbing for better error messages.
llama-cpp-bindings/src/{ggml_time_us,llama_time_us,max_devices,mlock_supported,mmap_supported}.rs	New focused utility modules + basic tests/docs.
llama-cpp-bindings/src/test_model.rs	Adds download_file_from() helper used by new model tests.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 18 out of 25 changed files in this pull request and generated 6 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 21 out of 28 changed files in this pull request and generated 2 comments.

Comments suppressed due to low confidence (1)

llama-cpp-bindings/src/json_schema_to_grammar.rs:38

• When the FFI call fails and error_ptr is null (e.g., allocation failure), this returns a generic "unknown error" and drops the actual status code. Consider including the status (and/or mapping known non-OK statuses like LLAMA_RS_STATUS_ALLOCATION_FAILED) in the error message so callers can distinguish failure modes even without an exception string.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.