Refactor gpubench-core to delegate all benchmark execution, device detection, and listing directly to the C++ native engine via gpubench-sys, resolving all discrepancies between CLI and GUI results

Author: Soddentrough

gpubench-core/Cargo.toml

@@ -7,3 +7,4 @@ edition = "2024"
 ash = "0.38.0"
 clap = { version = "4.4", features = ["derive"] }
 opencl3 = "0.12.3"
+gpubench-sys = { path = "../gpubench-sys" }