Allow overwriting solutions

[LasNikas]

At first glance, this keyword does not really make sense.
Since one would achieve the same result with save_final_solution=true and save_times=[tspan[end]], right?
However, one would then not have regular backups of the results.
As I mention in the docs:

Useful for memory efficiency in large simulations where only
the final result matters. The difference to simply setting
save_final_solution=true is that this provides regular checkpoint
backups at each save interval.

[codecov]

Codecov Report

❌ Patch coverage is 80.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 89.63%. Comparing base (acaf860) to head (f17871a).
⚠️ Report is 6 commits behind head on main.

Files with missing lines	Patch %	Lines
src/io/write_vtk.jl	75.00%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1041   +/-   ##
=======================================
  Coverage   89.63%   89.63%           
=======================================
  Files         121      121           
  Lines        8763     8764    +1     
=======================================
+ Hits         7855     7856    +1     
  Misses        908      908           

Flag	Coverage Δ
total	89.63% <80.00%> (+<0.01%)	⬆️
unit	65.77% <40.00%> (+<0.01%)	⬆️

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.

[LasNikas]

/run-gpu-tests

[svchb]

Can you just use nothing here?

[svchb]

if you modify 'add_underscore_to_optional_postfix' and 'add_underscore_to_optional_prefix' to return '' for nothing as well you can keep the original code

[svchb]

Or this does already work? I am not sure.

[LasNikas]

Hmm but then it is not consistent with pvd = paraview_collection(collection_file; append=iter > 0).
What's wrong with -1?

[svchb]

But if you overwrite there shouldn't be a collection anyway?

[LasNikas]

Here, the collection consists only of the current solution. Is this wrong?
What is wrong with iter=-1?

[svchb]

No its not wrong but you could skip writing the collection file since with one timestep in it it doesn't really make sense.

[svchb]

@efaulhaber ?

[efaulhaber]

From a usage perspective, I think it doesn't make sense to have a PVD file if we already know there is only ever going to be one VTK file in the collection.

But I also don't think it makes sense to just omit the iteration number. Say we want to use a regular saving callback (with very large intervals) and one specifically for restarts (the purpose of this PR, no?). First issue would be that both generate fluid_1.pvd, so we definitely don't want a PVD from the overwriting callback. Second, we would get fluid_1_0.vtu, fluid_1_1.vtu etc. and fluid_1.pvd from the regular callback plus one fluid_1.vtu from the overwriting callback. This is confusing. I would prefer adding a suffix like "current" or so. Then we get fluid_1.pvd from the regular callback and fluid_1_current.vtu from this callback.