* add cuda KVCache kernel
* annotation benchmark_kvcache_copy
* add use cuda
* fix import path
* move benchmark scripts to example/
* rm benchmark codes in test_kv_cache_memcpy.py
* rm redundancy codes
* rm redundancy codes
* pr was modified according to the review
