Separate benchmarks for individual steps of `make match` and `make map` (computation, then collection of results + filtering)

[karel-brinda]

Currently, if the aggregation of results fails, no match benchmark is produced