Add a new `group` field to the storage flex-model, letting device entries
reference a power sensor that represents a group of devices (e.g. a hybrid
inverter shared by a battery and PV). The group sensor's own flex-model entry
constrains the group's aggregate power:

- power-capacity: hard constraint (both directions)
- consumption-capacity / production-capacity: soft constraints, enforced via
  breach commitments with default breach prices (10000 <currency>/kW)

The group's scheduled aggregate power is saved to the group sensor. Nested
groups are supported (cycles rejected); groups require multi-device
flex-models and members must share a commodity.

Implemented on top of the existing device-group solver machinery
(ems_constraint_groups from the multi-commodity work, and grouped
FlowCommitments from PR #1934); no changes to the optimizer itself.

Also fixes a latent device-index misalignment in MetaStorageScheduler._prepare,
where per-device lists were built from the full flex-model list (including
stock-only entries) instead of the filtered device models.

Closes #2092

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TAad46Ayg86DHpL6nY54sX

Add a new `group` field to the storage flex-model, letting device entries
reference a power sensor that represents a group of devices (e.g. a hybrid
inverter shared by a battery and PV). The group sensor's own flex-model entry
constrains the group's aggregate power:

- power-capacity: hard constraint (both directions)
- consumption-capacity / production-capacity: soft constraints, enforced via
  breach commitments with default breach prices (10000 <currency>/kW)

The group's scheduled aggregate power is saved to the group sensor. Nested
groups are supported (cycles rejected); groups require multi-device
flex-models and members must share a commodity.

Implemented on top of the existing device-group solver machinery
(ems_constraint_groups from the multi-commodity work, and grouped
FlowCommitments from PR #1934); no changes to the optimizer itself.

Also fixes a latent device-index misalignment in MetaStorageScheduler._prepare,
where per-device lists were built from the full flex-model list (including
stock-only entries) instead of the filtered device models.

Closes #2092

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TAad46Ayg86DHpL6nY54sX
Signed-off-by: F.N. Claessen <claessen@seita.nl>

… asset tree (#2277)

Extend the storage flex-model's group field to also accept {"asset": <id>},
so a group can be identified by an asset-keyed flex-model entry - the form
that DB-stored flex-models naturally produce. This makes intermediate power
constraints fully definable via an asset tree with stored flex-models, with
an empty flex-model in the scheduling trigger.

- Asset-referenced group entries define no power sensor of their own; the
  group's scheduled aggregate power is saved via the entry's consumption
  and/or production output sensors, following the usual output-sensor
  conventions (full profile on a single sensor; clip-split when both given).
- Sensor-referenced groups keep saving the aggregate to the group power
  sensor, and may now also define output sensors.
- The UI asset flex-model editor recommends the parent asset when editing a
  child's group field (one-click suggestion), and hints in both the parent's
  and children's editors when the parent's flex-model defines power-capacity.
- New tutorial: toy example for intermediate power constraints, driven
  entirely by DB-stored flex-models.

Bug fixes uncovered along the way:
- AssetTriggerSchema.check_flex_model_sensors raised KeyError on asset-only
  flex-model entries.
- The freeze_server_now test fixture leaked its server_now monkeypatch into
  all subsequently run tests in the same process, causing order-dependent
  duplicate-key failures when scheduling jobs saved beliefs with identical
  (frozen) belief times.


Claude-Session: https://claude.ai/code/session_01TAad46Ayg86DHpL6nY54sX

Signed-off-by: F.N. Claessen <claessen@seita.nl>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

Adapt the group feature to the typed DeviceInventory introduced in PR #2321:

- Classify group entries in the inventory (new DeviceRole.GROUP), as the
  first classification step, so a group entry's own power sensor or output
  sensors cannot make it pass for a device.
- Replace the scheduler's ad-hoc _group_models/_group_to_devices derivation
  with the inventory's group_entries field and group_to_devices resolution
  (transitive leaf-device resolution with cycle detection).
- Drop this branch's own copy of the stock-only misalignment fix
  (device_sensor_id_to_index and the device_models zip fix), which main
  now solves by construction via the inventory's canonical device indices.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015qxM7UZ5wHTz3ftz1Mf9yy
Signed-off-by: F.N. Claessen <felix@seita.nl>