Rework the maximum branch per seed cut

[stephenswat]

This commit reworks the CKF cut on the maximum number of branches per initial seed. The current implementation makes little sense as it resets its internal counter in every step, which means that we can actually create far more tracks than the cut would suggest. This is therefore a borderline bugfix.

[sonarqubecloud]

Quality Gate passed

Issues
1 New issue
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[stephenswat]

Performance summary

Here is a summary of the performance effects of this PR:

Graphical

Tabular

Kernel	68f95fc	7076463	Delta
fit_forward	52.15 ms	51.87 ms	-0.5%
propagate_to_next_surface	42.41 ms	42.58 ms	0.4%
fit_backward	26.48 ms	26.48 ms	-0.0%
count_triplets	8.33 ms	8.34 ms	0.0%
find_triplets	4.43 ms	4.44 ms	0.2%
find_tracks	3.73 ms	3.72 ms	-0.3%
count_doublets	1.72 ms	1.73 ms	0.8%
find_doublets	1.27 ms	1.27 ms	-0.1%
ccl_kernel	873.47 μs	871.56 μs	-0.2%
Thrust::sort	455.88 μs	456.30 μs	0.1%
fit_prelude	409.73 μs	410.42 μs	0.2%
remove_duplicates	388.52 μs	393.49 μs	1.3%
select_seeds	347.00 μs	347.15 μs	0.0%
build_tracks	148.06 μs	147.12 μs	-0.6%
update_triplet_weights	99.37 μs	99.50 μs	0.1%
apply_interaction	53.04 μs	53.64 μs	1.1%
fill_finding_propagation_sort_keys	35.13 μs	34.85 μs	-0.8%
estimate_track_params	34.36 μs	34.41 μs	0.1%
populate_grid	30.38 μs	30.45 μs	0.3%
count_grid_capacities	29.12 μs	29.09 μs	-0.1%
unknown	20.53 μs	20.46 μs	-0.3%
fill_finding_duplicate_removal_sort_keys	18.60 μs	18.55 μs	-0.3%
form_spacepoints	12.45 μs	12.46 μs	0.1%
reduce_triplet_counts	5.38 μs	5.40 μs	0.4%
fill_fitting_sort_keys	2.28 μs	2.29 μs	0.6%
make_barcode_sequence	1.01 μs	1.02 μs	0.8%
fill_prefix_sum	165.50 ns	165.39 ns	-0.1%
Total	143.50 ms	143.40 ms	-0.1%

Important

All metrics in this report are given as reciprocal throughput, not as wallclock runtime.

Note

This is an automated message produced on the explicit request of a human being.

[stephenswat]

Note that this is currently breaking our ITk reconstruction so this is quite urgent. ;)

[krasznaa]

Don't claim that I would fully, or even reasonably partly understand the changes. But they look generally reasonable.

As long as https://github.com/acts-project/traccc/blob/main/device/alpaka/src/finding/combinatorial_kalman_filter.hpp gets the same update as well, we can go ahead with this.