Retire 28 stale expected-red enrollments: the decl_facts cluster passes now

[gunbai-bot]

main is red on 28 identities that are green in substance.

#8439 narrowed the reflection fixture pool roots, dropping "dag/std" (99.0% of the bytes those roots reached). The 28 rows of the decl_facts cluster were enrolled as expected-red; before the narrowing they were BUDGET-REFUSED -- undecided, not failed -- and after it they execute and pass. The floor reports it directly:

required-floor: STALE-QUARANTINE <id> is enrolled as expected-red and PASSED
  -- remove it from v2.workflow.floor_expected_red

Measured, not assumed:

run	head	cluster rows
main pre-#8439	b7564be	28 BUDGET-REFUSED
main post-#8439	2b95d86	28 STALE-QUARANTINE

This removes exactly those 28 from floor_expected_red, by identity.

Scoped deliberately. The same run reports 5 further stale enrollments -- dissolution_census_witness_test, extdeps_scope_placement_gate_loudness_witness, root_d_checkpoint_scalar_declared_arity_witness, and two rust_test_fixtures_import_closure_witness rows. They are not this lane's: nothing here changed their cost or their verdict. Sweeping them in would hide five unrelated decisions inside one lane's cleanup, so they are reported and left.

The remaining 102 BUDGET-REFUSED identities on main are the separate #8423 un-masking and are untouched here.

— sent from sharp-raven-273

[gunbai-bot]

Closing this as superseded, not merging it: main already retired all 28.

#8437 (bddb7cbbfe5, merged after I cut this branch) rewrote the same roster region and removed every one of the 28 decl_facts identities along the way. Checked at identity grain rather than by diff shape: all 28 grep to 0 occurrences in origin/main:src/v2/workflow/floor_expected_red.dag, so the merge conflict here resolves to an empty change.

That also means the conflict was the signal, not the problem. Re-applying my removal on top of main removes nothing, and merging a no-op that claims to retire 28 enrollments would leave a commit message asserting work that main had already done.

My error, plainly: I opened this without first checking whether a fix was already in flight for the same red. The measurement in the PR body stands (the 28 rows moved BUDGET-REFUSED -> PASSED once #8439 narrowed the roots) and it is now recorded on a merged commit that is not mine.

Still unswept, and still not this lane's: the 5 non-cluster stale enrollments named in the body -- dissolution_census_witness_test, extdeps_scope_placement_gate_loudness_witness, root_d_checkpoint_scalar_declared_arity_witness, and two rust_test_fixtures_import_closure_witness rows. Whoever picks those up should confirm against current main first; #8437 may have taken some of them too.

— sent from sharp-raven-273