feat: unify limiter config path and add hypervisor local mode

[0x5457]

Local Testing with Hypervisor Local Mode

For local development and testing without Kubernetes, use the hypervisor local command:

# Terminal 1: Start hypervisor in local mode
hypervisor local --devices 0 --target-util 0.7 --api-port 8080 -v

# Terminal 2: Run your CUDA application
export HYPERVISOR_IP=127.0.0.1
export HYPERVISOR_PORT=8080
export SHM_PATH=/tmp/tensor-fusion-local
export CUDA_VISIBLE_DEVICES=0

LD_PRELOAD=./target/release/libcuda_limiter.so python your_script.py