Add configurable event loop factories to worker and scheduler CLI

[Davixk]

Is your feature request related to a problem?

TaskIQ creates event loops internally in both CLI execution paths:

• worker subprocesses call asyncio.new_event_loop()
• the scheduler calls asyncio.run()

Applications that require a specific event loop implementation cannot provide one. One example is async Psycopg on Windows, which cannot run on the default ProactorEventLoop and requires a selector-based loop.

A process-wide event loop policy can currently work around this, but asyncio policies are deprecated in Python 3.14 and scheduled for removal in Python 3.16. Hardcoding WindowsSelectorEventLoopPolicy, as proposed in #641, would therefore solve only the worker case and depend on an API that is already being removed.

Describe the solution you'd like

Add an optional --loop-factory module:callable option to both taskiq worker and taskiq scheduler.

The imported callable should return an asyncio.AbstractEventLoop.

For workers, the factory should be resolved inside each worker subprocess and take precedence over automatic uvloop selection. For the scheduler, it should be passed to the scheduler runner. Existing behavior should remain unchanged when the option is omitted.

Examples:

taskiq worker app:broker --loop-factory app.loops:selector_loop_factory
taskiq scheduler app:scheduler --loop-factory app.loops:selector_loop_factory

Describe alternatives you've considered

• Setting a process-wide asyncio event loop policy. This is deprecated in Python 3.14 and scheduled for removal in Python 3.16.
• Automatically forcing a selector policy on Windows. This changes the default globally, covers only one platform, and still relies on the deprecated policy API.
• Reimplementing TaskIQ's worker or scheduler launchers in application code. This duplicates TaskIQ's process supervision and CLI behavior.

Which component would this affect?

Configuration