Repoint hardware_selection to std.measure MoneyAmountMicro authority

[gunbai-bot]

Summary

• Repoint hardware_selection.dag and carrier witnesses (measure_magnitude_carrier_test, hardware_selection_witness_test) from product.compute_fabric money aliases to the pre-existing std.measure authority: MoneyAmountMicro, money_amount_micro, money_amount_micro_count (already at dsl/std/measure.dag:201-207 on main).
• std/measure.dag is unchanged in this diff (zero net lines).
• compute_fabric.dag and hetzner_cost_quote_witness_test.dag intentionally untouched — hub PR scope per connector migration sequencing (vivid-wren-440); removing MoneyMicros there is a separate slice.

Scope note (§3)

This PR closes the hardware_selection + carrier-test consumer fork only. The remaining MoneyMicros/money_micros definitions in compute_fabric.dag:123-130 and hetzner round-trip are tracked for the hub relocation PR, not nicknamed anew here.

Test plan

• CI floor batch 1: dsl_compile_clean_gate_passes
• CI floor batch 2: discovery corpus (1111 witnesses)
• cargo test -p v1-compiler-tests measure_field_access

[gunbai-bot]

All three findings were valid against an earlier commit but are already addressed on head 3876d1a8f0:

1. compute_fabric.dag fork — Intentionally untouched per hub serial plan (vivid-wren-440). This PR only repoints scoped consumers (hardware_selection, carrier/witness tests) to the existing std.measure authority. compute_fabric deletion is deferred to the connector remodel PR. Diff vs main: 0 lines changed in compute_fabric.dag.

2. MoneyMicros alias in std/measure.dag — Removed. Diff vs main: 0 lines changed in std/measure.dag. Callers now use the canonical MoneyAmountMicro / money_amount_micro / money_amount_micro_count (already present at lines 201–208).

3. Duration leak in std/measure.dag — Removed (was out of scope). No Duration addition remains in this PR.

Current PR scope (3 files only):

• dsl/product/hardware_selection.dag
• dsl/test/claim/measure_magnitude_carrier_test.dag
• dsl/test/claim/hardware_selection_witness_test.dag

Local verification: gunbc compile --source-root dsl --source-dir dsl --target rust → 623 files emitted, 0 diagnostics.

— sent from neat-cat-88

[gunbai-bot]

Thanks for the detailed §3 read. Addressing each finding against current HEAD (b13d5e4072); gh pr diff 5912 --name-only touches exactly 3 files and does not include dsl/std/measure.dag.

1. Parallel authority / nicknaming (MoneyMicros vs MoneyAmountMicro)

Partially valid concern, out of scope for this slice.

• This PR does not add MoneyAmountMicro / money_amount_micro / money_amount_micro_count — those already live on main at dsl/std/measure.dag:201-207 (e.g. extdeps/cloud/hetzner.dag already imports them).
• What this PR does: repoint the hardware_selection + carrier-test consumers that previously imported MoneyMicros/money_micros from product.compute_fabric to the existing std authority.
• The surviving MoneyMicros/money_micros in compute_fabric.dag:123-130 and the hetzner round-trip (money_micros(count: money_amount_micro_count(...))) are intentionally deferred to the hub connector PR (vivid-wren-440 sequencing). Parent explicitly scoped this PR to hardware_selection + carrier tests only; touching compute_fabric.dag or hetzner here would race the hub remodel.
• Net: this PR reduces the fork (3 consumers moved off compute_fabric) without minting a second std alias. Full single-authority consolidation completes when the hub PR deletes the compute_fabric definitions.

2. Title/body overclaims Duration

Already corrected. Title is now "Repoint hardware_selection to std.measure MoneyAmountMicro authority". PR body updated to remove Duration/MoneyMicros-addition claims. Duration dual-home (compute_fabric.dag:121 vs std/types.dag:148) is hub/out-of-scope here.

3. Naming regression (MoneyAmountMicro vs MoneyMicros)

Disagree for this PR; manager decision.

Parent slice spec explicitly: "Do NOT add MoneyMicros to std/measure.dag" — consume the existing MoneyAmountMicro carrier already grounded next to MoneyAmount<Micro>. Promoting MoneyMicros as the std alias is a separate naming/consolidation decision tied to the hub PR that removes compute_fabric's duplicate. Renaming std's established MoneyAmountMicro (used by hetzner extdeps catalog rows) in this consumer-only PR would widen blast radius without closing the compute_fabric fork.

No code change requested — findings 1 and 3 describe debt correctly but belong to the hub relocation slice; finding 2 is addressed via title/body correction.

— sent from neat-cat-88