Allow automatic grouping of hits based on a group_distance when creating an Event

[fhagemann]

Following the discussion on #460, this PR provides a method to automatically group hits in subarrays based on their distances, such that charge clouds close to each other can interact, but the interactions between charge clouds far apart can still be neglected.

@apmypb Thanks for taking care of the core implementation of group_points_by_distance using breadth-first search, which was optimized to have little allocations and fast run times.

To stick to the same example as in #460, where two points are 1mm apart, see the results for passing no group_distance (the previous default), group_distance = 0.5mm (where the charge clouds are not grouped) and group_distance = 2mm (where the charge clouds are grouped).

using SolidStateDetectors
using Unitful
using Plots
T = Float32

sim = Simulation{T}(SSD_examples[:InvertedCoax])
simulate!(sim)
pos = [CartesianPoint{T}(0.02,0,0.05), CartesianPoint{T}(0.02,0.0,0.051)]
E = [1u"GeV", 1u"GeV"]
N = 5

@btime evt1 = Event($pos, $E, $N)
evt1 = Event(pos, E, N)
simulate!(evt1, sim, Δt = 1u"ns", max_nsteps = 5000, self_repulsion = true)
p1 = plot(sim.detector)
plot!(p1, evt1.drift_paths)

@btime Event($pos, $E, $N, group_distance = 0.5u"mm")
evt2 = Event(pos, E, N, group_distance = 0.5u"mm")
simulate!(evt2, sim, Δt = 1u"ns", max_nsteps = 5000, self_repulsion = true)
p2 = plot(sim.detector)
plot!(p2, evt2.drift_paths)

@btime Event($pos, $E, $N, group_distance = 2u"mm")
evt3 = Event(pos, E, N, group_distance = 2u"mm")
simulate!(evt3, sim, Δt = 1u"ns", max_nsteps = 5000, self_repulsion = true)
p3 = plot(sim.detector)
plot!(p3, evt3.drift_paths)
    
plot(p1, p2, p3, size = (1500,500), layout = (1,3), legend = false)

  11.870 μs (180 allocations: 7.20 KiB)
  16.472 μs (257 allocations: 10.20 KiB)
  17.873 μs (242 allocations: 9.72 KiB)

The grouping results in some extra runtime when creating the Event, but can help saving time when calculating self_repulsion later in the pipeline. If self_repulsion = false, I would still recommend to use the method without group_distance to achieve the previous results and run times.

[fhagemann]

Note: this PR is non-breaking as it defaults to the old implementation

[fhagemann]

The other example from #460 also works:

using SolidStateDetectors
using Unitful
using Plots
T = Float32

sim = Simulation{T}(SSD_examples[:InvertedCoax])
simulate!(sim)

pos = CartesianPoint{T}(0.02,0,0.05)
E = T(1)u"GeV"

evt0 = Event([pos], [E], 50)
simulate!(evt0, sim, Δt = 1u"ns", max_nsteps = 5000, self_repulsion = true)

evt1 = Event([pos, pos], [E/2, E/2], 50)
simulate!(evt1, sim, Δt = 1u"ns", max_nsteps = 5000, self_repulsion = true)

evt2 = Event([pos, pos], [E/2, E/2], 50, group_distance = 1u"mm")
simulate!(evt2, sim, Δt = 1u"ns", max_nsteps = 5000, self_repulsion = true)

plot(evt0.waveforms[1],  label = "One hit")
plot!(evt1.waveforms[1], label = "Two hits, independent")
plot!(evt2.waveforms[1], label = "Two hits, interacting")

[fhagemann]

I am still not quite convinced by the name group_distance. Maybe this could become something like interaction_distance or max_interaction_distance?
@oschulz any suggestions?