Add lossy RLGC ladder model to transmission line element

[esaruoho]

Summary

Adds optional series resistance (R) and shunt conductance (G) to TransLineElm. When both are zero the element is the original lossless delay line. When either is non-zero, the element switches to an N-segment RLGC lumped-element ladder, giving distributed attenuation and dispersion.

Supersedes #309. That PR started as a simple two-port resistance approximation, which @pfalstad correctly noted didn't model real lossy lines (no dispersion, instant attenuation). I reworked it into the RLGC ladder mid-PR but the conflicting comments and history made it hard to review. Opening this as a clean PR with one focused commit.

What's new

Parameter	Behavior
Delay, Impedance (existing)	Unchanged
Resistance (new, total Ω)	Series R distributed across segments
Conductance (new, total S)	Shunt G distributed across segments

Internally:

• L_total = Z0 · delay, C_total = delay / Z0 derived from existing parameters — no new electrical inputs other than R and G.
• N segments auto-selected in [4, 50] based on delay/timestep so the ladder resolves the propagation correctly at the current simulation timestep.
• Each segment uses trapezoidal companion models for L and C, matching InductorElm / CapacitorElm.
• Series R and inductor companion resistance combined into a single Norton equivalent per segment to minimise added internal nodes.
• R == 0 && G == 0 → falls back to the original delay-line code, no overhead.

Addresses both points from #309

@pfalstad concern	How RLGC ladder addresses it
"Not showing attenuation over the length of the line" — at R=500 attenuation was instant	Losses accumulate per segment, visible in the voltage coloring along the line
"Real lossy line has dispersion and this model doesn't show that"	The LC ladder naturally produces frequency-dependent propagation; pulses broaden as they travel

A proper SPICE LTRA (lossy transmission line, distributed) implementation would still be more accurate for high-frequency / high-Q work, but the lumped RLGC ladder is a substantial step up from the previous two-port resistance and is the standard educational/intermediate model.

Test plan

• R = G = 0 (default for old circuits): identical to the existing lossless delay line.
• R = 100 Ω: signal attenuates progressively along the line; voltage coloring shows the gradient.
• Step input on a lossy line: rise time visibly broadens at the far end (dispersion).
• Reflection from open/short termination: amplitudes attenuated by round-trip loss.
• Save/load: R and G round-trip via XML attrs; old circuits load with R=G=0.
• Vary timestep: N adapts so the visible behaviour stays consistent.

🤖 Generated with Claude Code