Feature: tighter timing

[aarondcohen]

I started using your library as a benchmarking replacement for matcha. I noticed some timing strangeness where the run times would fluctuate wildly depending on test order. This should address some of those timing discrepancies:

• only use low resolution timer for test time threshold
• use high resolution timer for each timed block
• remove everything beyond the loop control and the timed code from the timed block
• run calculations on summation of timed block durations

Also some project hygiene:

• update dependencies
• clean up whitespace