docs(test): harden deferred OpenDSS geometry against silent defaults

[BurhanAbdullah]

Summary

This draft contributes a regression harness and a concrete safety contract for the geometry-normalization issue in #479.

The current reader can normalize Line objects that reference the deferred OpenDSS geometry family using the Line class factory impedance defaults, while also deriving conductor count from the balanced phase default. The result can be electrically wrong before any writer is invoked.

Included

• Regression harness for a geometry-backed overhead line.
• Regression harness for a single-conductor SWER geometry.
• Explicit parse-time-diagnostic expectation.
• Documentation of the fail-closed contract and the required long-term canonical representation.

Why draft

The canonical DistLine currently requires a concrete linecode and impedance matrices. A complete fix therefore needs an intentional representation for a deferred/unresolved line rather than replacing the impedance with a sentinel that could accidentally reach a solver. I have deliberately not introduced a speculative schema change in this PR.

The target invariant is:

Unsupported geometry must remain source-retained and explicitly diagnosed; it must never silently acquire OpenDSS factory impedance or a fabricated conductor count.

Explicit linecode= paths remain unchanged.

Closes the implementation gap described by #479 once the deferred-line representation is agreed.

[samtalki]

@BurhanAbdullah thank you so much for your contribution. :)
I will get to this asap!