Reorganize benchmarks/dbscan

[aprokop]

The current code mixes different clustering algorithms in a single place. It has many issues:

• unclear parameters for each algorithm
• difficulty in adding new algorithms
• hard to read code

The main motivation for this cleanup is to prepare for adding a distributed dbscan benchmark.

This PR:

• Splits dbscan benchmark into 3 (dbscan, hdbscan, mst)
• Refactors code
• Removes unused data converter

[dalg24]

I would prefer if we moved the common instantiation things into a library to speed up compilation.

[aprokop]

I would prefer if we moved the common instantiation things into a library to speed up compilation.

Can you please elaborate?

What I'd really like is to be able to specify in CMakeLists.txt which instantiations to do, and not having to edit the source files switch statements. Then, I would just instantiate it for a couple dimensions by default. When needing more, like a paper, I think it would be OK to allow longer compilation times.

[dalg24]

We are compiling the same thing multiple times, one for each exe

[aprokop]

We are compiling the same thing multiple times, one for each exe

Which thing?

[dalg24]

Minor comments.
Looks good after you fix them.

[dalg24]

Suggested change

	target_link_libraries(ArborX_Benchmark_DBSCAN.exe ArborX::ArborX Boost::program_options cluster_benchmark_helpers)
	target_link_libraries(ArborX_Benchmark_DBSCAN.exe PRIVATE ArborX::ArborX Boost::program_options cluster_benchmark_helpers)

[aprokop]

Doesn't really matter for the executable. We don't do this anywhere else.

[aprokop]

Couple builds did not start (HIP, SYCL) but they passed before clang-tidy check, so that's fine.