Consider alternative benchmark groupings

[dgeb]

The current groupings evolved a bit awkwardly as different benchmarking scenarios were added to the mix.

These groupings could be refactored to more natural categories, such as:

• Language:

  • JS
  • WASM - emscripten-generated

• Thread spawning / teardown:

  • Spawn / teardown threads for each test
  • Reuse pre-spawned threads

• Algorithm:

  • Fibonacci(50)
  • Simple multiplication
  • Quicksort

Work location:

• On main thread only
• On each background thread
• On main thread + each background thread
• On main thread, invoked once from each background thread
• On main thread, invoked for each iteration

• Work iterations

  • min
  • max
  • step

• Background threads

  • min
  • max

Although these groupings are arguably more clearly divided, they would lead to a larger matrix of results.