Tests should include some benchmarking

[ToxicFrog]

I've talked a lot about implementing things like an optional C backend to speed up vstruct, but how slow is it really? I need a way to consistently test performance so that I can tell what's actually making it slow, and whether the things I think are speeding it up really are.

At the moment the approach is to switch to ss1edit and time how long it takes to load ARCHIVE.RES, but that's not really a good solution.

[ToxicFrog]

fa84229 added some basic performance testing, but I want something more rigorous (and easier to use for regression testing) before I close this.