Fix: Remove duplicate idx_tasks_session_id from SCHEMA_SQL

Issue: #28617

Root Cause

When the kanban schema gains a new indexed column (session_id in the tasks table), the CREATE INDEX IF NOT EXISTS statement inside SCHEMA_SQL references the new column — but on existing databases, that column doesn't exist yet. SQLite throws OperationalError: no such column: session_id during conn.executescript(SCHEMA_SQL) in connect(), before _migrate_add_optional_columns() gets a chance to add the column.

IF NOT EXISTS only skips if the index exists — SQLite still validates the column reference. On an existing DB without session_id, this fails before the migration code runs.

Fix

Remove the CREATE INDEX IF NOT EXISTS idx_tasks_session_id ON tasks(session_id); from SCHEMA_SQL. The migration function _migrate_add_optional_columns() already adds both the column and the index correctly for legacy databases.

This is consistent with how run_id on task_events is handled — the idx_events_run index lives only in the migration function, not in SCHEMA_SQL.

Verification

• idx_tasks_session_id now only appears once, in _migrate_add_optional_columns() (line ~1176)
• idx_tasks_session_id no longer appears in SCHEMA_SQL
• Fresh installs: unaffected (migration runs, creates the index)
• Legacy installs: migration now runs successfully without crashing