Repair trace overlaps using geometry-sized clearance detours

[Abse2001]

When an overlap remains after force-based repair, fixed-distance doglegs can miss the same conflict at another size or across several trace vertices. Add a geometry-sized same-layer detour that uses the blocker extent, trace width, and declared clearance. It covers trace-to-pad, trace-to-plated-hole, trace-to-trace, and trace-to-via conflicts, including rotated obstacles.

The detour grows across same-layer vertices while preserving ports, layer transitions, through-obstacle spans, and other routes. It is attempted after existing candidates stall, without a low-DRC-count gate. Each stalled error is attempted once until the routes change. Acceptance requires fewer full-board DRC errors and no new DRC error identities. The repair does not change fabrication constraints or remove requested routes.

This is broader overlap handling, not a guarantee that every board is routable. Crowded multi-obstacle paths and terminals without a legal escape can remain unresolved. The original SRJ33 sample051 has not been demonstrated clean under its original fabrication rules. The existing explicitly permitted small-via test remains a separate configuration test.

Validation on c51c84722093839f9f603d26f5a78e8f05c9af91:

• Remote repair suite: 100 passed, zero failed; typecheck passed.
• 384 candidate configurations vary overlap kind, dimensions, 45-degree rotations, translation, route reversal, and detour direction. Segmented traces are checked with both the repair engine and the autorouter's exact @tscircuit/checks@0.0.163 version, added only as a test dependency.
• An eight-overlap board verifies that large blockers are repaired without changing endpoints or via rules.
• Remote run: https://github.com/tscircuit/high-density-repair03/actions/runs/34626335012

The existing terminal-via relocation overlap fix is retained, with its permission and drill-fit checks. This PR is stacked on the exact Repair03 dependency pinned by the autorouter baseline (5f6c9af547dbb70c8948227011e1769b5dfa16a5).

Applied autorouter PR: tscircuit/tscircuit-autorouter#2545. Full applied tests and unchanged-input same-machine benchmarks are pending. No tests, builds, installs, or solver benchmarks run locally.

[tscircuitbot]

Benchmark This PR

Run benchmarks by commenting on this PR:

/benchmark [benchmark.sh args...]
/benchmark [scenario-limit|all] --dataset <drc14|srj18> --concurrency <n> --effort <n> --max-iterations <n>

Everything after /benchmark is forwarded directly to ./benchmark.sh.

Examples:

• /benchmark -> run default SRJ18 benchmark settings
• /benchmark 10 -> run first 10 SRJ18 samples
• /benchmark --dataset srj18 -> run all SRJ18 samples
• /benchmark 1 --dataset srj18 -> run the first SRJ18 sample
• /benchmark --dataset drc14 -> run all DRC14 samples
• /benchmark all --concurrency 4 -> run all SRJ18 samples with 4 workers
• /benchmark --scenario-limit all --effort 2 -> run all samples with effort 2
• /benchmark --scenario-limit 20 --max-iterations 100 -> run 20 samples with a max iteration override

Any PR whose title contains [BENCHMARK TEST] will automatically run the benchmark workflow on PR updates.