Adds the following metrics:

[rkuris]

• each ffi batch reports "firewood.ffi.batch_ms" for elapsed millis
• each ffi batch reports "firewood.ffi.batch" for each run (allows the mean time per call to be batch_ms/batch)
• each predictive_reader reports "firewood.io.read_ms" for elapsed millis
• each predictive_reader reports "firewood.io.read" for each run

These should complete the following requested statistics:

• What percentage of time is spent performing blocking DB reads during EVM execution (now in firewood.io.read*)
• What percentage of time is spent during merkalization after EVM execution -> (now in firewood.ffi.batch*)

These other statistics were already available:

• What percentage of time is spent during EVM execution (the EVM records these)
• What is our CPU utilization during the benchmark (Are we CPU bound?) (the system level node recorder records these)
• What is our RAM utilization during the benchmark (Are we using memory more efficiently?) (the system level node recorder records these)
• What is our disk utilization during the benchmark (Are we fully utilizing our available IOPS) (the node recorder records these too)