Advisories 5.4: delete the GUNBC_REALIZE_ADVISORY logger and realize_advisory_soundness

[gunbai-bot]

Step 5, item 4 of the advisory program (node adhoc-e84e492c-5ef).

Dead by consumer search at origin/main: no workflow, hook, script or .dag emission sets GUNBC_REALIZE_ADVISORY. Its only readers were two opt-in if env::var(..).is_ok() gates. v2.workflow.realize_advisory_soundness was imported only by its own test module; no observation rows ever fed it, and the runs that would have produced them were cut.

Consumers removed (§3c):

• v1_compiler.cli_run emit_realize_advisory_for_rows and its private helper realize_advisory_optional_int (now unreferenced).
• The env-gated call sites in bin/claim_batch.rs and cli_run::required_floor_runner.
• The doc-comment pointer to the env var in derived_realization_schedule::derive_discovery_schedule_width.
• The emit_realize_advisory_for_rows DeclarationRef row in gunbc.interpreter_eval_correctness_seed_growth, since the declaration no longer exists.
• v2.workflow.realize_advisory_soundness and v2.test.claim.realize_advisory_soundness_test (six witnesses). These are declared as six WitnessDeleted arms in floor_grandfathered_removals, and the frozen cut is untouched, per that roster's contract.
• A stale prose pointer to the soundness module in v2.compiler.self_host.native_routing_frontier.

Kept: std.realize_pack realize_advisory. It is live through derived_realization_schedule::derive_discovery_schedule_width, which the required floor runner calls, and this PR does not touch it. The soundness witnesses guarded a law with no producer, so no real class loses evidence (§4b(4)).

v1-maintenance purpose test (gunbc.v1_maintenance_standing): this is pure subtraction of an unreached seed path. It removes host-Rust surface the self-host program would otherwise have to account for, and adds no v1 growth. cli_run is hand-authored seed host Rust, not a generated mirror, so no regeneration is involved.

Checks: cargo clippy -p v1-compiler --all-targets -- -D warnings passes (remote). The .dag side is checked by the CI floor. 00_compile.dag and 04_infer.dag are untouched.

🤖 Generated with Claude Code

[gunbai-bot]

On the open question in review 74299: std.realize_pack realize_advisory still has a live consumer after this PR. derived_realization_schedule::derive_discovery_schedule_width calls it through the interpreter bridge (run_in_context_with_args(.., "realize_advisory", ..)), and cli_run::required_floor_runner calls that function on every floor run to derive the discovery batch width. The PR body states this under Kept. So no §3c follow-up is owed.

— sent from tidy-wolf-843