Clean up how we log + time

[jonkeane]

We currently do a few hand-coded timing of steps. But maybe we should be more systematic and make a timing helper that we can use to "just" wrap specific functions (rather than starting a timer then stopping at the top|bottom of each)

[joosthooz]

The timing code is a relic from back when we were still experimenting and comparing performance, we could also just remove it.

[jonkeane]

Yeah, removing it is fine too, I guess really we would want to truly profile code like that when experimenting with speedups (e.g. confirming that our partitioning in dbgen works well, when we tackle tpc-ds, etc.)