Add shifting for InFlow-buffer particles

[LasNikas]

I ran a simulation with a step in the transition zone (see image below). This step induced periodic gaps in the buffer zone which didn't decay. Indeed, the opposite is the case. The simulation is becoming unstable.

So the workaround is to shift not only the fluid particles but also the buffer particles.
I only shift particles that are in the vicinity of the transition plane.

Left: No shifting of buffer particle. Right: Shifted buffer particles

[codecov]

Codecov Report

Attention: Patch coverage is 2.12766% with 46 lines in your changes missing coverage. Please review.

Project coverage is 70.22%. Comparing base (114a57b) to head (683b09c).

Files with missing lines	Patch %	Lines
src/callbacks/particle_shifting.jl	0.00%	44 Missing ⚠️
src/schemes/boundary/open_boundary/system.jl	0.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #768      +/-   ##
==========================================
- Coverage   70.60%   70.22%   -0.38%     
==========================================
  Files         106      106              
  Lines        6769     6804      +35     
==========================================
- Hits         4779     4778       -1     
- Misses       1990     2026      +36     

Flag	Coverage Δ
unit	70.22% <2.12%> (-0.38%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[efaulhaber]

Can you just cut off the shifting at a point? Is this stable? Or should this be a smooth transition from shifting to no shifting?

[efaulhaber]

1. This is not public API.
2. This is not available for every NHS implementation.

I see two options:

1. Use foreach_point_neighbor and skip all particle-neighbor pairs with particle outside the max distance (benchmark performance and compare with this function here).
2. If this is not fast enough, we can make foreach_neighbor public API.

[LasNikas]

The first option is not an option, because their exists no NHS for OpenBoundarySPHSystems.
Buffer particles don't need to interact with other particles or with itself.
I like the second option

[efaulhaber]

Buffer particles don't need to interact with other particles or with itself.

You are just implementing an interaction between buffer particles and other particles.

You can't just use another NHS and assume it works for a different set of particles. This is what PointNeighbors.requires_update is for.

[LasNikas]

Can you just cut off the shifting at a point? Is this stable? Or should this be a smooth transition from shifting to no shifting?

Yes, because these buffer particles don't calculate their physical properties physically but extrapolate (Tafuni) them or determine them with a model (Lastiwka). Here, the shifting acts basically as a rearrangement of the position of the particles.