Seperate benchmark data generation and benchmark explorer

[vchuravy]

Folsk should be able to use BenchmarkExplorer for their own datasets, for now our "early" users should be

• Trixi
• Enzyme
• KernelAbstractions

What do they need to do to generate benchmark results that are compatible with BenchmarkExplorer.

[vchuravy]

@yaraslauprogj right now Project.toml contains the data sources, that should not be the case.

We need to seperate the concerns of data generation (running the benchmarks) and data visualization.