Explain how coverage statistics are gathered. about coverage HOT 1 OPEN

There are two components to the coverage package:

1. bin/collect_coverage.dart: connects to a running VM or Dartium instance and, for each isolates gathers a hitmap (map of line number to hitcount) for each script loaded by that isolate. These hitmaps are then emitted to stdout in JSON format. Alternatively, the same JSON can be generated by running the VM with --coverage_dir=somedir.
2. bin/format_coverage.dart: takes the JSON generated by either method in the step above, merges it down to a single set of hitmaps, one per script file, and emits it in the specified format: either pretty-printed source marked up with a per-line hitcount or in LCOV format.

Last I looked, Coveralls relies on --coverage_dir which only works for VM coverage. I'm looking at options for automating collection of both VM and web coverage, which will likely require Dartium content-shell to be installed. #48 tracks that work.

I'll add documentation to README.md.

from coverage.