`RTO.connect` timeout docstring contradicts the default, and the `ResourceManager` is discarded

[sequoiap]

pyrolab/drivers/scopes/rohdeschwarz.py:110-121:

def connect(
    self, address: str = "", hislip: bool = False, timeout: float = 1e3
) -> bool:
    """
    timeout : int, optional
        The device response timeout in milliseconds (default 1 ms).
        Pass ``None`` for infinite timeout.
    """
    rm = visa.ResourceManager()
    if hislip:
        self.device = rm.open_resource(f"TCPIP::{address}::hislip0")
    else:
        self.device = rm.open_resource(f"TCPIP::{address}")

1. Docstring is wrong. 1e3 ms is 1 second, not 1 ms. A 1 ms VISA timeout would fail on essentially every operation, so a reader trying to raise the timeout from its apparent value will be confused about what units are in play.
2. timeout is typed float but documented int.
3. The ResourceManager is a local. pyvisa expects the ResourceManager to outlive the resources it opens; dropping the only reference leaves self.device depending on pyvisa's internal caching to stay valid. Store it as self._rm and close it in close().

The module docstring's "Common Issues" note already describes a failure mode where the scope is left "without data and with a bad connection" after a timeout -- worth checking whether the discarded ResourceManager contributes to that.

---

Found in a full-codebase audit at v0.4.0 (commit 1ce3146).