HAND lens E: retire transport_script_position_project.rs — lens .dag owns verdict

[briansrls]

Summary

First file in the HAND lens E dissolution queue (Chunk D): retire transport_script_position_project.rs so the interpreter evaluates v2/lens/host_language_transport_script.dag tables directly.

• Add transport_script_position_facts_for_path host builtin (fact extraction only) in module_path_index.rs
• Wire transport_script_live_literal_violation_count_for_path through pure .dag logic (transport_script_literal_violation_count(facts: …))
• Remove transport_script_literal_violation_count_for_path builtin and delete the projection .rs
• Drop from HAND_MAINTAINED_STAGE0_FILES, 05_emit_rust.dag mod list, and bootstrap copy/exclude lists
• regen_verify_gate_passes green on prior CI run (before fleet OOM kill)

Test plan

• regen_verify_gate_passes passed in CI batch 5 @ 9c14c21 (before OOM kill in batch 7)
• Floor discovery corpus 1120/1120 witnesses passed @ 9c14c21
• Fix duplicate HashMap import compile error from d7ea7d2
• CI re-trigger @ dff489f (prior failures were exit-137 OOM on srv1 runners, not code)

[gunbai-bot]

Addressing REQUEST_CHANGES (head 7bfe72ad8):

Duplicated dissolution-trigger literal — fixed. gunbc.hand_lens_host_bridge_scaffold_index is now the single authority; both hand_lens_host_bridge_scaffold_watchdog and v2.lens.host_language_transport_script import it. The watchdog roster check now requires exact equality to that constant (replacing the fail-open substring self-check).

ci_floor_plan zero-budget witness — the prior hunk was a CI unblock for #6059 (gunbc_ci_plan_spawn_width now int_min(floor, exec)); testing plan_spawn_width(0) was the wrong target. The witness now reads gunbc_ci_floor_spawn_width_for_budget(byte_size(0)) and still pins the exact conservative fallback (== 4), matching witness_floor_unreadable_budget_uses_conservative_envelope. Kept on this branch because floor CI was red without it; isolated in its own commit.

— sent from still-crab-325