Remove multiplet number estimator and use the counter result to resize the multiplet container

[beomki-yeo]

This PR addresses #93.
The size of multiplet (doublet, triplet, and seed) container used to be resized based on the empirical values obtained by cpu seed finding. The changes in this PR makes seed finding no longer rely on the empirical values but use the counting result to resize the container.

To make it possible, the counter objects had to be extended to include the more information on the number of multiplets.

Good thing is that the overall speedup gets improved because the size of doublet and triplet container is not over-estimated anymore:

Cuda seeding Elapsed time for 20 events
Before (1.4 sec) -> After (1.0 sec)

[krasznaa]

Formally I think this looks fine.

I was very confused by the doublet_per_bin and triplet_per_bin structs at first. Since those are really just an unnecessary complication over simply using an unsigned int by itself, like the code has done before.

I see that the interface must've been developed for the dublet_counter_per_bin struct at first, where a struct is actually meaningful. But... Is this uniform interface for these 3 structs actually important at any point? Did you write some generic code somewhere that would use these 3 types in a uniform way? I can't see that in the code... (Though I might just be missing it.)

As always, C++20 concepts would fit this type of code beautifully. Hopefully we'll get those in not too long...

[beomki-yeo]

Did you write some generic code somewhere that would use these 3 types in a uniform way?

Yeah. They have uniform structure to use get_ref_num() in cuda/utils/cuda_helper.cuh :p

[krasznaa]

Once #108 is merged in, this should still be updated to include those changes first. That should also give the CI a chance to fix the (hopefully temporary) hiccup that occurred in the last run.

But apart from these technicalities, I'm on board with this PR as well.

[beomki-yeo]

I want to merge this - but why review-required: main is not satisfied although it was already approved? @paulgessinger

[paulgessinger]

The review through the bot gets invalidated when you add new commits to the branch that are not just update commits.

[paulgessinger]

Should be green now

[beomki-yeo]

Thanks!