Fail closed on deferred OpenDSS geometry

[samtalki]

Why

OpenDSS lines whose impedance comes from LineGeometry, LineSpacing,
WireData, CNData, or TSData currently fall through to the Line class
factory impedance. That can turn a real geometry into plausible but wrong
electrical parameters and can turn a one-conductor SWER line into three
conductors without a parse diagnostic (#479).

This implementation is intentionally stacked on Burhan Abdullah's regression
and safety-contract PR #480. The temporary base branch is an exact mirror of
#480's head so this PR contains only the follow-up implementation commit. Merge
#480 first; this PR can then be retargeted to main without copying the
contributor's work.

What changed

• Represents deferred line impedance explicitly as DistLine.linecode: None.
• Replays phases, linecode, geometry, spacing, and conductor/cable-list
  assignments in source order so terminal maps retain the stated conductor
  shape.
• Emits a parse-time error diagnostic when geometry impedance has not been
  calculated.
• Preserves exact same-format DSS output from retained source.
• Omits unresolved lines from semantic DSS, PMD, and BMOPF output with an error
  diagnostic instead of inventing a linecode or electrical values.
• Keeps explicit linecode and inline-impedance behavior unchanged.

Validation

• cargo test -p powerio-dist
• cargo check -p powerio-dist --all-features
• cargo clippy -p powerio-dist --all-targets -- -D warnings
• cargo fmt --all --check
• git diff --check

The regression coverage includes four-conductor geometry, single-conductor
SWER, spacing plus conductor lists, parse severity, exact DSS echo, and every
semantic distribution writer.

Depends on #480. Addresses #479 without implementing the separate Carson/line
constants work tracked in #84.

[github-actions]

Conversion Matrix

Matrix generation failed before writing the report. See the workflow logs.

[BurhanAbdullah]

I traced the workspace compile failure after DistLine.linecode became Option<String>. The core direction in this PR is right; the remaining issue is propagation of the new unresolved state into downstream consumers.

The current failures are in facade code that still assumes linecode: String, notably powerio/src/stored/legacy09/document.rs and powerio/src/transform.rs (calls such as net.linecode(&line.linecode), eq_ignore_ascii_case(&line.linecode), and formatting line.linecode with {}). These should not unwrap/default None, because that would reintroduce the silent-fabrication hazard this PR is fixing.

Suggested invariant for each consumer:

let Some(linecode) = line.linecode.as_deref() else {
    // record/return the existing unresolved-impedance error for this semantic path
    // and do not lower/serialize/analyze this line
    continue; // or return the appropriate readiness/lowering error
};

Then use linecode for lookup/diagnostic formatting. In readiness/lowering code, None should make the network not electrically lowerable rather than appearing as an unknown named linecode.

I also searched the tree for remaining direct line.linecode assumptions. Besides the two compile-failing facade paths, powerio-matrix/src/matrix/multiconductor.rs and some tests still use the old mandatory-string shape on current main. Worth auditing those as part of the same API propagation so a future all-workspace build cannot accidentally accept unresolved geometry.

Happy to keep validating the propagation against #480's fail-closed contract; the important point is preserving None as a first-class unresolved electrical state, never coercing it to a placeholder.