Don't plumb an executor through when bootstrapping tools.

Review Request #1634 - Created Jan. 21, 2015 and submitted

Benjy Weinberger
jsirois, patricklaw, zundel
- There was only one case where we were actually using this, which was
  when IvyResolve bootstraps xalan, and uses its own nailgun instead of
  a subprocess JVM like all the other tools. This seems arbitrary.

- A tool gets bootstrapped only once after a clean-all, and there are only
  a handful of tools, so there's no need to use a nailgun.  The complexity
  of a nailgun is only worth it for things that are going to run a lot
  even without cleaning, such as the java compiler.

- This simplifies the code slightly, ahead of a larger simplification I'd
  like to do to the ivy/nailgun/jvmtask triad.

CI passes except for an unrelated test that is broken in master for some weird reason:

Eric Ayers
Patrick Lawson
Benjy Weinberger
Review request changed

Status: Closed (submitted)

Change Summary:

Submitted as ea923e322ede524f9b8da29450df17d0be679b0c.