Persist Ingested Data Into Postgres #36

Problem

There is external data being scraped but not being populated in the database.
Refer to #5, #7, #8, #9

Goal

Every ingestion source persists into its Postgres table through a shared repository layer, so downstream model training
and the API read from one source of truth.

Scope

• Build out foresight/repository/ with one repository per table (weather, events, competitor_rates,
  flight_arrivals, and hotel tables).
• Define a single persistence contract: adapter produces normalized records → repository upserts
  them.
• Add idempotency/upsert support
• Wire external adapters to persist through the repository as the reference
  implementations.
• Consider database best practices (cascading, indexing, cardinality, primary and foreign keys)

Acceptance criteria

• External source tables are populated by running their ingestion locally
  - weather, events and the three hotel tables: done. flight_arrivals is
  wired and tested but needs OpenSky and RapidAPI credentials.
  competitor_rates has no adapter and no source ticket.
• Re-running the same ingestion produces zero duplicate rows (upsert verified)
• All DB writes go through repository layer. No adapter touches a SQLAlchemy session or
  SQL directly