High memory usage on large tasks

[trifle]

Hi,
thanks for making tourbillon-celery!

I have an (admittedly rare) use case where some of my tasks are in the range of 2-10 MB large.
This means that tourbillon will consume a large amount of RAM (I suspect up to 100GB, ouch!).

A brief investigation leads me to suspect that the reason for this is a LRU cache in celery's events.State class. State has a tasks attribute that is defined as:

https://github.com/celery/celery/blob/82e478e07cb9beee490369b71b7008fcf0f08f89/celery/events/state.py#L377

    self.tasks = (LRUCache(max_tasks_in_memory)
                  if tasks is None else tasks)

with the default parameter in its __init__ method set to max_tasks_in_memory=10000.

So - it would be very nice if one could tune this parameter.

[trifle]

By the way, this is exactly what flower (celery monitoring web app) does:

https://github.com/mher/flower/blob/a86bf0478f23f5834b72c256ee0ce0afc4199fed/flower/app.py#L42