[test] fuzz and/or unit-test under resource tracker to confirm calibrations

[graydon]

We have a test that checks actual resource use vs. modeled, but we currently run it only on a small set of cases. This should be expanded to work with either a fuzzer or (parts of) the unit testsuite, or both.

[graydon]

Pushing this back to P1 and moving to next milestone; we likely aren't going to have time for it before final. Worst case we're a bit wrong, and we'll need to do a vote to update the on-chain fee schedule.