feat: add GitHub Copilot SDK backend with CLI fallback - #2392

Compared 00efd6e7969837ae4a9f11d8d504dcd3b20b09df (base) against a1c8ab6e173a7b91d7d975aad57c6850f42c55b1 (head).

• graphify/cli\.py · dispatch_command: unsupported — not verifiable: all 9 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly SystemExit — names the real obstacle, not a sampling gap)
• graphify/dedup\.py · _llm_tiebreak: unsupported — not verifiable: all 105 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly TypeError — names the real obstacle, not a sampling gap)
• graphify/llm\.py · _build_image_refs: unsupported — parameter `root` is annotated `Path` — outside the synthesizable primitive/collection set
• graphify/llm\.py · _call_llm: unsupported — not verifiable: all 9 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly TypeError — names the real obstacle, not a sampling gap)
• graphify/llm\.py · _default_model_for_backend: unsupported — not verifiable: all 9 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly KeyError — names the real obstacle, not a sampling gap)
• graphify/llm\.py · _image_notes: unsupported — non-vacuity: domain too small (only 1 distinct inputs exercised, need 3) — 'no divergence' would be near-vacuous
• graphify/llm\.py · _read_files: unsupported — parameter `units` is annotated `'list[Path | FileSlice]'` — outside the synthesizable primitive/collection set
• graphify/llm\.py · extract_corpus_parallel: unsupported — parameter `root` is annotated `Path` — outside the synthesizable primitive/collection set
• graphify/llm\.py · extract_files_direct: unsupported — parameter `root` is annotated `Path` — outside the synthesizable primitive/collection set
• graphify/llm\.py · label_communities: unsupported — not verifiable: all 12 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly TypeError — names the real obstacle, not a sampling gap)

Backend availability notices:

• 5 changed test function(s) excluded from verification (a test is the oracle, not the subject)
• none of the 0 changed id(s) matched a code node in the graph (0 known) — 19 changed function(s) recovered directly from the git diff instead (graph may be stale/unsynced; consider a rebuild)

Formal verification is advisory: equivalent is a proof over a bounded sound subset; may_equivalent is sampled, not proven; unsupported and error are honest abstentions.