Add actionable algebraic-loop diagnostics

[jamestjsp]

Closes #167.

What changed

• add AlgebraicLoopError with implicated named channels and the actual feedthrough-equation condition number
• preserve errors.Is(err, ErrAlgebraicLoop) and support errors.As through Connect, ConnectByName, LFT, Feedback, and zero-delay approximation failures
• identify named input/output channels from the singular or nearly singular SVD mode only after a named interconnection has failed
• keep valid dynamic feedback through zero-feedthrough models accepted
• preserve the wrapped algebraic-loop error in the consolidated Feedback delay path from PR Fold SafeFeedback into Feedback delay options #166

Semantics

Condition is the condition number, not a reciprocal estimate. Exact singularity reports +Inf; numerically rejected but nonsingular equations report a large finite value. The previously suggested scalar gain 0.999 is not near-singular under this relative condition metric, so the regression uses a genuinely ill-conditioned MIMO feedthrough equation.

Operations without signal metadata return the typed condition diagnostic with an empty Signals slice. The structural DiscretizeWithOpts rejection remains the sentinel fallback because it does not factor a feedthrough equation and therefore has no honest condition estimate.

Validation

• go fix ./...
• go vet ./...
• go test -v -count=1
• go test -race -count=1 ./...
• before/after benchmarks for Feedback, approximated-delay feedback, ZeroDelayApprox, and Connect; allocations are unchanged

Rebase

Rebased onto the current main after PR #166. The renamed conflict site retains this PRs %w wrapping and algebraic-loop diagnostic behavior.