fix: restore go1 reset domain randomization path

[TATP-233]

Summary

• remove the custom Go1WalkTask.reset() override so Go1 uses the shared NpEnv -> DomainRandomizationManager reset flow again
• add a dedicated MuJoCo slow test that verifies Go1 reset applies base mass and COM domain randomization to the backend state

Issue

Closes #173

Validation

• UV_CACHE_DIR=.uv-cache make check
• UV_CACHE_DIR=.uv-cache uv run ruff check src/unilab/envs/locomotion/go1/joystick.py tests/envs/test_go1_domain_randomization.py
• UV_CACHE_DIR=.uv-cache uv run pyright tests/envs/test_go1_domain_randomization.py
• UV_CACHE_DIR=.uv-cache uv run pytest tests/envs/test_go1_domain_randomization.py tests/envs/test_env_configs.py -q -k "go1_mujoco_reset_applies_base_mass_and_com_domain_randomization or go2_mujoco_reset_applies_kp_kd_domain_randomization"
• UV_CACHE_DIR=.uv-cache uv run pytest tests/integration/test_reward_injection_integration.py -q
• UV_CACHE_DIR=.uv-cache uv run pytest tests/config/test_config_system.py -q -k "go1_joystick or Go1JoystickFlatTerrain"

Known local failures

• UV_CACHE_DIR=.uv-cache uv run pre-commit run --all-files still fails on existing repo-wide pyright / mypy issues outside this change (for example benchmark/ and several pre-existing test typing paths).
• UV_CACHE_DIR=.uv-cache make test aborts while collecting tests/algos/test_mlx_ppo.py due a local MLX runtime crash.
• UV_CACHE_DIR=.uv-cache uv run python scripts/train_rsl_rl.py task=go1_joystick training.sim_backend=mujoco algo.max_iterations=1 also aborts in the same local MLX runtime path before reaching Go1 training execution.