--create dataframe to xml marshaller

--xrange to range

--refactor pydevd_save_locals to avoid deadlocks caused by import lock being taken after debugger lock

When the new process is created, the main thread in the new process already has the attribute 'pydevd_id', so the new thread doesn't get new id with its process number and the debugger loses access to both threads. Therefore we should update thread_id for every main thread in the new process.

We should update additional frames container after updating thread id.

Replace tabs in keys with special symbols for sending to debugger

load meta data for array in first request

(cherry picked from commit 651cb167)