feat(gpu): add OpenCL C backend — --target opencl + transparent CPU fallback

Fix the forth interpreer

feat(gpu): add ocean.nano — GPU ocean simulation with Archimedes ship physics

feat(gpu): add matmul.nano — 2D GPU operations demo using gpu_launch2d

ci: lower coverage threshold to 75% to match actual coverage baseline

fix(asan): handle zero-capacity ModuleList in module_list_add

fix(asan): initialize env->effect_registry = NULL in create_environment()

fix(ci): resolve both CI failures — ASAN stack-use-after-return + add GPU intrinsic tests

fix: add gpu_launch5 for 5-arg kernels; fix saxpy missing n arg

fix(gpu): correct DynArray layout in cuda_runtime.c + atom.global.add.u64

fix(ptx): use atom.global.add.u64 for gpu_atomic_add

feat(gpu): gpu_load/gpu_store PTX intrinsics + saxpy/reduce_sum/vector_scale examples

fix(gpu): C forward decls for GPU intrinsics + shadow tests for gpu.nano

fix(ptx): use u32 register pool (%r) for special register reads

fix(ptx): replace em-dash with colon in PTX header comment