Add automatic schema versioning with startup migration

On startup, the app now checks if the database schema version matches
the code. If there's a mismatch or no version exists, it automatically
runs a full data migration before starting.

- Add backend/version.py with SCHEMA_VERSION constant
- Add backend/migration.py with extracted migration logic
- Add SchemaVersion model to track DB version
- Add version check functions to database.py
- Update app.py lifespan to use check_and_migrate_if_needed()
- Simplify migrate_csv_to_db.py to use shared logic

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

backend/models.py

@@ -3,9 +3,11 @@ SQLAlchemy database models for school data.
 Normalized schema with separate tables for schools and yearly results.
 """
 
+from datetime import datetime
+
 from sqlalchemy import (
     Column, Integer, String, Float, ForeignKey, Index, UniqueConstraint,
-    Text, Boolean
+    Text, Boolean, DateTime
 )
 from sqlalchemy.orm import relationship
 from .database import Base
@@ -133,6 +135,21 @@ class SchoolResult(Base):
         return f"<SchoolResult(school_id={self.school_id}, year={self.year})>"
 
 
+class SchemaVersion(Base):
+    """
+    Tracks database schema version for automatic migrations.
+    Single-row table that stores the current schema version.
+    """
+    __tablename__ = "schema_version"
+
+    id = Column(Integer, primary_key=True, default=1)
+    version = Column(Integer, nullable=False)
+    migrated_at = Column(DateTime, default=datetime.utcnow, onupdate=datetime.utcnow)
+
+    def __repr__(self):
+        return f"<SchemaVersion(version={self.version}, migrated_at={self.migrated_at})>"
+
+
 # Mapping from CSV columns to model fields
 SCHOOL_FIELD_MAPPING = {
     'urn': 'urn',