vmarea and heap exit flags not reset on re-init

[s-kanev]

This came up in the following case:

• static attach with a client (drmemtrace), which detaches with dr_app_stop_and_cleanup().
• later, from the same process, invoke raw2trace_t, which uses dr_init_standalone() to map modules and decode instructions.
  Mapping modules then fails because update_dynamo_areas_on_release() doesn't remove VM areas thinking it's exiting, because both vm_areas_exited and heap_exiting are still set.

The fix is simple, don't statically initialize them, but use the proper _init() functions.

[derekbruening]

Re-initializing the same DR library instance is not fully supported yet: #2157 covers the main feature and initial work for full (non-standalone) re-attach.

[s-kanev]

Understood, I'm guessing the standalone case is much simpler than #2157 because it touches on fewer static variables.