Add performance metrics for block devices usage

I wonder if we should try creating tool for monitoring block I/O operations triggered in particular thread.

In theory it should be possible to create a Ruby gem with native extensions that would use ptrace() call to track disk operations during execution of code inside the block.

Metrics::Block.trace do
  do_something
end

In theory we should be also able to track fork() and execve() and attach to subsequent threads created inside the block.

In practice, we would need a PoC to see if this actually doable. Would it be worth to investigate this?

What do you think @yorickpeterse @pcarranza @stanhu?