Restructure orchestration models into typed profiles and tasks

[daniel-slaugh]

Summary

Replace the generic Task model (with task_type discriminator) and DataConnection model with per-type profile and task tables. Remove OrchestrationSystem entirely — the internal Celery worker is the only orchestration runner.

Backend

• Remove OrchestrationSystem model and all references
• Keep DataConnection as the ETL profile (holds extractor/transformer/loader config + notification recipients)
• Add new models: AggregationProfile, AggregationTask, AggregationMapping (with typed fields: source_datastream FK, target_datastream FK, statistic, timezone_mode, timezone_offset)
• Add stub models for DerivedProfile, DerivedTask, DerivedMapping, DerivedMappingSource
• Add stub models for QualityProfile, QualityTask
• Migrate existing Task rows into ETLTask or AggregationTask based on task_type
• Separate TaskRun tables per type (or shared with discriminator — TBD)
• Add per-type Celery task functions (run_etl_task, run_aggregation_task, etc.)
• Add per-type API endpoints with strongly typed schemas
• Update send_orchestration_notifications to iterate all profile types
• Update hydroserverpy client with typed namespaces (hs.etl.connections, hs.aggregation.profiles, hs.aggregation.tasks, etc.)

Frontend

• Redesign orchestration page: profiles as the primary grouping, type-specific task table columns
• Type-specific task creation forms with datastream pickers (replace UUID text fields)
• Stepped creation flow (profile → schedule → mappings)
• Profile detail/edit slide-over (view details, edit, add task — all as explicit buttons)
• Empty state cards for new workspaces explaining each orchestration type
• Type filter in toolbar replacing the data connection toggle panel

References

• Derived Datastream Orchestration #368 (Derived datastream orchestration)
• Support Automated Data Quality Management #373 (Automated data quality management)