Commit the deployed Owned stack sources and broadcast record

[mfw78]

The mainnet Owned stack was deployed from uncommitted working-tree sources. Commit them verbatim so develop matches what is live on chain, with a bytecode-reproducibility gate.

Why

develop (d1f3603) contains none of the Owned stack that was broadcast to mainnet: no Owned.sol, no Commitment.sol, no deploy script, and the modified OrderModule/OrderDescriptor and type sources exist only in a dirty tree. #35 promised a canonical deployment record for monitoring services, yet the deployed code has no tracker footprint at all. Everything fork-consuming, including shepherd's #398 train, is gated on a committed, reproducible source of truth.

Scope

• Commit verbatim from the deployed working tree: src/types/Owned.sol, src/libraries/Commitment.sol, script/deploy_OwnedStack.s.sol, modified src/OrderModule.sol and src/OrderDescriptor.sol, the four type sources (GoodAfterTime, StopLoss, PerpetualStableSwap, twap/TWAP), and the modified tests.
• Commit broadcast/deploy_OwnedStack.s.sol/1/run-latest.json unmodified as the deployment record.
• Gate G0: forge build output vs deployed runtime bytecode modulo metadata (or Blockscout re-verification) for both CREATE2 addresses.

Done when

• CI is green on develop with the committed sources.
• Bytecode reproducibility holds for OwnedTWAP 0x4E17A65D14e7f37D2A9F0389F17eFd41aAa64c91 and the registry 0xf9ba6F64c9b41Df1cEe76A50e2039D3847064232 (modulo metadata, or Blockscout re-verify passes).
• A test asserts the descriptor digest is still 0xb499…ce1a.
• The broadcast record is committed unmodified.