Failed discovery returns None with no status code, log line, or exception

[tipatterson-dev]

Sub-issue of #46.

Problem

Node.discover_systems() (node.py:316-338) returns None when the HTTP request fails:

if result.ok:
    ...
    return new_systems
else:
    return None

No log line, no exception, no status code — the response object is discarded. A caller sees None and has no way to learn whether the server was down, rejected its credentials, or returned a 500. Worse, the falsy return is easy to conflate with "this node has no systems" (which returns []), so the common for s in node.discover_systems() or []: idiom turns an auth failure into a silent zero-iteration loop.

This is the read-side twin of #42: same "failure becomes indistinguishable from an empty success" shape, on GET rather than POST.

Reproduce

node = Node(protocol="http", address="localhost", port=8282,
            username="wrong", password="wrong")
print(node.discover_systems())   # None — no indication it was a 401

Point at a stopped server for the same result via a different cause.

Possible fix

Two defensible options; pick one and apply it consistently to every discovery method:

1. Raise (consistent with what System.insert_self() silently swallows failed POSTs, causing downstream AttributeError #42 established for the POST side) — raise ResourceDiscoveryError(...) carrying status code and body, once Typed exception hierarchy instead of bare Exception #47 lands the hierarchy.
2. Log and keep returning None — lower-churn, preserves the documented return contract (the docstring already says "or None if the HTTP request failed"), but still leaves callers unable to react programmatically.

At minimum the failure must be logged at ERROR with the status code and response text, so it isn't invisible. Worth auditing System.discover_datastreams() / discover_controlstreams() at the same time — they call res.json() without checking res.ok at all, so a failed listing raises a JSON decode error rather than anything meaningful.