Fix n-bodies stack-overflow error with new type Bodies

[b-fg]

New AutoBodies type is implemented that can solve the n-bodies problem reported in #88. It holds a vector of type AutoBody and implements superposition of sdf and map functions by iteration instead of recursion.

Tests show same results as the recursive implementation of the superposition, while n-bodies can be handled now without hitting the stackoverflow error. Also, the measure function is much faster than before when superposing more than 5 bodies. The cost of the iterative method does not really scale when adding more bodies, while the recursive method seems to scale exponentially.

Here are some test when adding number of bodies n=6 and n=11 (max. I can handle in my computer with recursion)

radius = rand(n)
circles = [(x,t) -> √sum(abs2,x)-r for r ∈ radius]
body = AutoBody(circles[1])
for i ∈ 2:n
    body += AutoBody(circles[i])
end
bodies = AutoBodies([AutoBody(c) for c ∈ circles], +)
xy = rand(2)

n=6

@btime measure($body, $xy, 1.)
# 5.436 μs (25 allocations: 22.44 KiB)
# 3.670 μs (57 allocations: 2.05 KiB)

n=11

@btime measure($bodies, $xy, 1.)
# 549.630 μs (25 allocations: 4.96 MiB)
# 3.735 μs (63 allocations: 2.20 KiB)

I do not particularly like that there is currently AutoBody and AutoBodies, but I did not want to get rid/break the previous implementation. Maybe this can be merged in a nicer way though.

[codecov]

Codecov Report

Attention: 3 lines in your changes are missing coverage. Please review.

Comparison is base (3bbbca1) 96.55% compared to head (fa2bd1d) 96.12%.

Files	Patch %	Lines
src/AutoBody.jl	90.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #95      +/-   ##
==========================================
- Coverage   96.55%   96.12%   -0.44%     
==========================================
  Files          12       12              
  Lines         465      490      +25     
==========================================
+ Hits          449      471      +22     
- Misses         16       19       +3     

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

[b-fg]

Sadly I cannot test on MacOS, so I am not sure. But the fix on that test is because it fails in Julia 1.10.0 for any OS.

Other than this, the PR is mostly finished. The main new features of this PR is that Bodies can account for many AutoBody and perform measure without maxing out the stack. Also, individual body forces can now be extracted using

force = WaterLily.∮nds(sim.flow.p,sim.flow.f,sim.body.bodies[i],t)

[b-fg]

Ready for review.

[b-fg]

Should we merge this or just close it? @weymouth

[weymouth]

We should merge it.