fix: field bus forwarder CIM profile, argument order, and mrid option (GPY-003)

[craigpnnl]

Summary

Three related fixes to the field bus proxy forwarder, all surfaced during the GPY-003 investigation.

• FieldProxyForwarder imported cimgraph.data_profile.cimhub_ufls unconditionally at module scope. That profile no longer exists in the pinned cim-graph release, so any import of the module raised ModuleNotFoundError. Replaced it with a configurable cim_profile constructor parameter (default cimhub_2023), resolved dynamically via importlib.import_module, mirroring the pattern already used in agents.py. Verified cimhub_2023 exposes both DistributionArea and Substation, which is everything the constructor needs.
• The CLI in forwarder.py called FieldProxyForwarder(username, password, connection_url, None) against a constructor signature of (connection_url, username, password, mrid). Username landed in the connection_url slot, and no mrid was ever passed, so substation discovery could not work even once the import bug above is fixed. The call now passes each value by keyword into the correct slot.
• The CLI had no way to supply an mrid at all. Added a --mrid click option (with a GRIDAPPSD_FIELD_BUS_MRID environment fallback, matching the existing option style) and threaded it into the corrected constructor call.

Added a unit smoke test that imports the forwarder module and asserts the resolved default CIM profile exposes DistributionArea and Substation, so a future profile default swap that regresses this cannot merge silently. The test does not instantiate FieldProxyForwarder and does not stand up any broker or Blazegraph instance, since the constructor connects to live infrastructure.

Test plan

• New test test_field_proxy_forwarder_cim_profile.py passes
• Full field-bus test suite passes (excluding the Docker-dependent Artemis test, which needs a live container)
• start-field-bus-forwarder --help shows the new --mrid option
• No version fields touched (pixi.toml, both pyproject.toml files)
• Working tree clean; the compute_req_log.txt side effect from importing gridappsd (tracked separately as GPY-004) was restored, not committed