Fix duplicate recurring enqueues when multiple schedulers run

[knifecake]

Summary

• fix recurring execution recording to use create semantics instead of update_or_create
• raise/catch RecurringExecution.AlreadyRecorded to skip duplicate runs
• keep enqueue + recurring execution record in one transaction so duplicate inserts rollback
• add regression test for same-task/same-run_at double enqueue
• add CHANGELOG entry under Unreleased

Why

When two scheduler instances enqueue the same recurring task at the same run_at, update_or_create allowed both jobs to be enqueued and the second worker only updated the record. This caused duplicate job execution.

This aligns behavior with Solid Queue's AlreadyRecorded flow.

Closes #52

Validation

• uv run python manage.py test tests.test_recurring_execution --keepdb
• uv run python manage.py test tests.test_recurring_task tests.test_recurring_execution --keepdb
• uv run ruff check steady_queue/models/recurring_execution.py steady_queue/models/recurring_task.py tests/test_recurring_execution.py