v8.6.7

The L2FT Transfer Templates sheet rendered empty in the demo
because no row in the seeded data carried a non-NULL
template_name. Two gaps:

1. The M.3.10g TransferTemplatePlant picker only handled
   templates whose first leg_rails entry was a TwoLegRail;
   single-leg first leg_rails (the only shape both shipped L2
   instances declare — MerchantSettlementCycle /
   InternalTransferCycle) got skipped with an Omitted
   reason. Net: 0 TransferTemplatePlant rows on either
   spec_example or sasquatch_pr.
2. Even when plants exist, the demo CLI helper
   (cli/_helpers.py::build_full_seed_sql) called
   default_scenario_for(instance) with the default
   l1_invariants mode — and M.4.2a re-categorized
   transfer_template_plants as broad-mode-only. So even after
   fix (1), TT plants never made it into the SQL the demo applies.

Fix:

• auto_scenario.py — picker branches on the first leg_rails
  entry's rail kind. TwoLegRail keeps the existing
  source/destination_role resolution; SingleLegRail resolves
  leg_role and reuses the same account_id for both
  source_account_id and destination_account_id on the
  plant (the emit ignores the destination for single-leg).
• seed.py::_emit_transfer_template_rows — branches on the
  rail kind. TwoLegRail keeps the 2-leg debit+credit shape.
  SingleLegRail emits ONE leg with direction per
  rail.leg_direction (Variable treated as Debit for
  plant purposes; closing-leg semantics aren't material to
  surfacing data on the explorer). chain_children attach via
  transfer_parent_id regardless of parent rail shape.
• cli/_helpers.py::build_full_seed_sql switches to
  mode="l1_plus_broad" so the demo gets BOTH the L1
  SHOULD-violation plants AND the broad-layer TT + RailFiringPlant
  rows. The L2FT Rails sheet now also surfaces broad-mode
  rail firings on top of the baseline.

Single-leg TT firings surface on the L2FT Template Instances
table as Imbalanced against expected_net = 0 (one bare
leg can't sum to zero) — accurate L1 representation. Multi-leg
shared-transfer cycles (sibling legs joining via the same
transfer_id by transfer_key) are still deferred.

Hash impact: L1-invariants-mode YAML seed_hash is unchanged
(picker change is gated by broad-mode include). Broad-mode and
l1_plus_broad-mode sidecar hashes (_BROAD_MODE_HASHES in
tests/data/test_l2_seed_contract.py) re-baked.