Profiling the compiler

So far very little work has gone into making the compiler fast. For the most parts I thinks it's "fast enough", but since there presumably are no big projects written in Carp this is hard to know for sure. I think that profiling the compiler to at least know what parts are slow could be an interesting and worthwhile endeavour. This issue can be a good place for reporting the results and discussing possible improvements.

Start by identifying the Carp compiler entry point and running a profiler against a representative compilation workload. The issue names no files, tests, workload, or profiling method, so those need to be established before work can be scoped. Done would include recorded profiling results and a documented discussion of the slow areas and possible improvements.

Written by the indexing model from the issue text.