Task: Write the swap legs on import and record their direction

Make the swap legs the importer builds reach the database, keyed by the trade and the ordinal as bond_legs is, with the leg's direction recorded.

• ores_trading_swap_legs_tbl is keyed by (tenant_id, trade_id, leg_number) with no surrogate id, and carries payer.
• swap_instrument_mapper::map_leg reads legData/Payer.
• The import writes every leg of a swap-shaped trade after its instrument, and fails the item when a leg write fails.
• The component regenerates drift-clean and builds, and the codegen suite passes.

The importer already builds the legs and stamp_ids already stamps them, so the work is two small changes: the leg gains the direction ORE requires, and the import writes the legs it already builds.

1. payer joins swap_leg, map_leg reads legData/Payer and reverse_leg emits it, so the direction survives import and export both.
2. The handler writes each leg after its instrument, and fails the item when a leg write fails. The leg's id is minted at the write site, because the table still keys the leg by its surrogate.

The re-key to (trade_id, leg_number) was attempted and set aside. It surfaces a codegen defect: the compound-key path cannot generate a compilable eventing integration test, because cpp_nats_integration_test.cpp.mustache inserts each extra key column's raw domain value where the generated service takes every key part as text. Fixing that template belongs with the codegen work, so the re-key was recorded as a capture and the surrogate key stood here. The codegen defect is fixed and the re-key landed in Re-key the swap legs to the trade.

The amount and rate children that carry an amortising notional and a stepping rate are the next unit: map_leg reads only the first notional, rate and spread today, so the children would be empty until the mapper reads the lists.

The database is recreated rather than migrated (recreate_is_the_only_db_migration).

Every number this task claims about a tree, with the command that produced it and the commit it was measured at.

The swap_leg row keeps its :default: 1 syntax on leg_number rather than the :default_value: its siblings use. It emits the correct SQL default, so it is left alone here rather than changed in an unrelated diff.

Manual QA scenarios (scaffolded via compass add test_scenario) that verify this task. Link new ones here as they're created; the scenario doc itself links back via its "Verifies task" field.

The swap legs the importer builds now reach the database, and a leg records which side pays.

• The direction is recorded both ways. payer joins swap_leg, map_leg reads ORE's required legData/Payer and reverse_leg emits it. Before this, even a mapper that built the legs dropped the direction, and the stored shape had nowhere to put it.
• The import writes the legs. The handler's swap_instrument_data branch wrote the instrument and its call dates and dropped the leg vector that stamp_ids had already stamped with the trade and the activity. It now writes each leg after its instrument and fails the item when a leg write fails, so a stored swap is an instrument rather than a row. This is finding R15 of the family design, and it is why a swap imported before this change exported with no legs at all.
• The re-key was attempted and set aside. Dropping the surrogate id for (trade_id, leg_number) fails to compile a generated test: the compound-key path inserts each extra key column's raw domain value where the generated service takes every key part as text, so an int ordinal lands in a std::vector<std::string>. The fix is a template change that belongs with the codegen work, so the re-key was recorded as a capture and the surrogate key stood here. The codegen defect is fixed and the re-key landed in Re-key the swap legs to the trade.

Verification: the component regenerates with no drift; the codegen suite is 1276 passed; the full preset build exits 0 with no errors and no warnings.

Not verified live: that an imported swap lands its leg rows. The database and the service fleet are down in this environment. Proving it means compass db recreate followed by a fleet start and an ORE import, which is a deliberate, whole-environment step rather than something to do as a side-effect of a build.