Fix regression in set_timer, old events not removed

[ankith26]

Fixes #2763

This was a minor regression introduced in my older event refactor PRs that got merged in 2.0.1

Setting a timer for an event should discard older timers for the same event. While this behavior was not previously documented, this was the standard behaviour for a long time until 2.0.1 when my changes inadvertently changed behaviour here. (Funnily, I had documented this behaviour in the same PR I broke it 😅, I remember having tested it, but it somehow worked back then due to some test weirdness/randomness in the broken behaviour)

While having multiple timers for the same event running independently sounds like a good idea, this fix is needed for backwards compatibility of older code using this function.

This PR also updates the tests for this function.

Also what do y'all think of a keyword argument for this function, something like remove_old_timer=True, which can be set to false if someone wants to place in an event on the timer and not remove the older one?

[illume]

Thanks for this. Especially for adding those tests :)