In storage we now use single task for injector (we look it up by name)
and we restore result mapping for that task in constructor if there
is one in flow detail we are given. This has visible effect
of injected data staying in storage after resumption.
Closes-bug: #1240931
Change-Id: I2e8fb2c4b5cca769d36f578a934119db4a530f5c