Add 128-bit integer plugin (nautilus-int128) with MLIR intrinsics, runtime, docs, and tests

[PhilippGrulich]

Motivation

• Provide a traceable signed 128-bit integer type for Nautilus to allow arithmetic, bitwise, shifts, comparisons, casts, and memory operations on platforms with __int128 support.
• Expose a plugin that lets scalar backends use portable runtime helpers while the MLIR backend emits native i128 operations for better performance.
• Add automated tests and an optional libFuzzer differential harness to validate correctness across backends.

Description

• Add ENABLE_INT128_PLUGIN CMake option and wire plugins/int128 into the top-level build when enabled.
• Introduce plugins/int128 library implementing runtime helpers (int128.cpp) and a public header include/nautilus/int128.hpp exposing nautilus::int128 with constructors, operators, loads/stores, and accessors.
• Add MLIR intrinsic plugin (MLIRInt128Intrinsics.cpp/.hpp) to replace runtime calls with native i128 LLVM/MLIR operations when the MLIR backend is enabled.
• Include installation rules and a public CMake alias target nautilus::nautilus-int128 in the plugin CMakeLists.
• Add documentation docs/int128-plugin.md describing semantics, memory behavior, MLIR interaction, and the fuzz target behavior.
• Add execution tests (plugins/int128/test/Int128ExecutionTest.cpp) using Catch2 and an optional differential fuzz target (plugins/int128/test/fuzz/Int128Fuzz.cpp) built when ENABLE_FUZZING=ON.

Testing

• Ran the plugin unit tests via Catch2 (nautilus-int128-tests) through ctest and all tests passed.
• Built the project with MLIR enabled to exercise MLIRInt128Intrinsics registration and the build completed successfully in that configuration.
• Verified that the fuzz target nautilus-int128-fuzz is conditionally added when ENABLE_FUZZING=ON; the fuzz target builds with libFuzzer sanitizer flags when enabled.

---

Codex Task