Support for other forms of `$DOCKER_HOST` passing

There are two cases when `docker_ip` fixture returns incorrect value:
1. Non-default docker context used.
2. Remote docker used via `ssh://...` schema instead of `tcp://`.

I have an idea:
```python
import pytest_docker, json, re, shlex, socket

@pytest.fixture(scope="session")
def docker_ip() -> str:
# These two statements supports all forms of passing dockerd endpoint to docker (ezcept -H option):
# 1. DOCKER_HOST variable
# 2. DOCKER_CONTEXT variable
# 3. docker context use
# 4. Defaults to unix://...
context_name = (
(pytest_docker.plugin.execute("docker context show") or b"default")
.decode("utf-8")
.strip()
)

# This block is almost same as in origin
docker_host = pytest_docker.plugin.execute(
'docker context inspect -f "{{.Endpoints.docker.Host}}" '
+ shlex.quote(context_name)
).decode("utf-8")
if not docker_host or docker_host.startswith("unix://"):
return "127.0.0.1"
match = re.match(r"^tcp://(.+?):\d+$", docker_host)
if match:
return match.group(1)

# Last-chance failover for non-tcp remote docker daemons
info = json.loads(pytest_docker.plugin.execute("docker info -f json"))
if info["Name"] == os.environ.get("HOSTNAME"):
return "127.0.0.1"
return info.get("Swarm",{}).get("NodeAddr") or socket.gethostbyname(info["Name"])
```

Other options are much worse:
1. Parse `~/.ssh/config` to detect chain of `ProxyJump` and `ProxyCommand`.
2. Use something like `execute("docker run --rm alpine ip route get 8.8.8.8")` which will not work in private network.
3. Require swam-enabled daemons and force `docker info -f {{.Swarm.NodeAddr}}`.

Maybe somebody knows better way to discover connectable address of remote docker daemon, but note that it can be non-connectable at all if ssh goes throw firewall or if it has `--ip=127.0.0.1` in parameters.

BTW: One can configure proxy of `tcp://` daemon (like Portainer does).

As a last resort user of some very special config can rewrite `docker_ip` fixture in `conftest.py`.