std.change: import any and Present so its own names resolve natively

[gunbai-bot]

Adds std.change imports of v2.std.algebra { any } and std.optional { Present }.

RED on main: native CLI (ctrl-build --remote, linux/amd64, one dispatch building //gunbc/instruments:v2-native-cli, then emit --entry std.change --source-root <dag copy> --source-root <src/v2 copy> over the 14-module import closure only): refuses at change.dag any with resolve_unbound_name_is_declared_elsewhere v2.std.algebra.any.
Positive: with the imports, std.change's own names resolve; next fatal is in another module (Absent, declared in several modules). Imported direct and pipe calls both bind (probe modules r1/r2 pass resolve; local r3/r4 too), so there is no resolver defect: my earlier 'import does not bind' reading was a harness bug (dag/std/algebra.dag and src/v2/std/algebra.dag collided under one root).
Residue unchanged otherwise (std.types, std.algebra); #13586 holds the receipt. Not merged.

🤖 Generated with Claude Code

Harness note: measure with TWO source roots (--source-root --source-root <copy of src/v2>). dag/std/algebra.dag and src/v2/std/algebra.dag are different modules (std.algebra vs v2.std.algebra) with the same relative path, so a single merged root overwrites one with the other and fabricates unbound-symbol refusals.

[gunbai-bot]

Added Absent to the std.optional import. Measured same setup (two roots): previous head refused at Absent, resolve_unbound_name_is_declared_in_several_modules (std.optional Optional.Absent, std.upsert_decision ObservationVerdict.Absent, v2.std.execution_surface Absent: distinct meanings, legitimate scoped reuse, not a nickname; the author must import, so the refusal is right). New head: first fatal moves to change.dag:246 'reverse' (kernel primitive; PR 2's class). — sent from lively-bear-908

[briansrls]

APPROVE at exact head 24f1a84fb9235acf7a3f1a8fa81ee90e613673f1. No blocking finding in the import repair.

The entire delta is two import lines in std.change: std.optional { Present, Absent } and v2.std.algebra { any }. I checked the producers and the consuming sites, not just the spelling. std.optional owns Optional<T> = Absent | Present { value: T }; std.keyed_row::keyed_row_find returns the optional keyed row and constructs that same Present. The Absent patterns in std.change eliminate those optional results (and the module's own optional-returning diff helper), not another module's unrelated Absent meaning. Importing both arms from this authority is correct.

The imported any is the ordinary v2.std.algebra::any<T>(xs: FreeMonoid<T>, predicate: fn(T)->Bool) with the existential fold behavior the piped membership check needs. No function body, assertion, refusal or return arm changes in this PR; no homonym-based fallback or extra exported declaration is introduced. The dependency is stated at the consumer, rather than teaching the resolver to guess which any or Absent it meant.

Scope: this repairs the named source bindings. It does not certify that std.change and its canonical dependency closure now pass all native resolve/infer/emit stages. The PR's copied-closure experiment and direct/pipe probes remain author-run evidence; I did not replay them. The body is slightly stale: this head imports Absent too, and the later primitive/none/type refusals should not be summarized as complete std.change success.

Exact-head workflow 37794779505 succeeded: seed, generated, floor, emit-build and witnesses passed; rust-unit-tests was skipped. The generated job passed all-target lint and the one-emission stage0 check. I did not independently reconstruct per-claim execution from a floor artifact or run a local compiler.

No further code change or new lane requested. Use the normal required checks on the composed merge revision. This approval does not approve stacked #13591. No merge or enqueue performed.

[gunbai-bot]

Superseded by #13641 (v1 closeout): this head is an ancestor of integration/v1-closeout.