performance measurements for type-safe bundling

[afoxman]

• [ ] Compare the cost of bundling with and without TypeScript validation
• [ ] Optimize where possible, using TS compiler flags, TS API caches, local file/snapshot caches
• [ ] Parallelize if necessary
• [ ] Consider using a persistent tsserver process to get the best caching and responsiveness.

The output of this issue is a report showing perf metrics of bundling across small, medium and large projects, both with and without type checking. Metrics include clock time, CPU time, peak memory, I/O (bytes read/written), and an interesting internal info V8 can expose to us.