Use a benchmarking framework for compactify benchmarks

[brookslogan]

The compactify benchmarks currently:

• can give "false" comparisons locally depending on what other tasks are running during the locf test vs. the no-locf test (maybe not an issue for the Action runner?)
• might penalize the second variant run due to heavier garbage collection load
• (might fall victim to other benchmarking gotchas)

We should try shifting to a pre-package benchmarking solution; Ken linked me this list.

Interacts with #237; microbenchmark has built-in visualizations.