chdir was causing long delays, moved the checker from this place to the connectivity check, added keepalive

Changed submission recovery to a better version that keeps in mind the submission time. Changed to sacct since the job could be finished. Also now the jobs are not cancelled if they're not duplicated.

(cherry picked from commit 2cb9eb1b)

(cherry picked from commit 01783457)

(cherry picked from commit 3deeaa46)

(cherry picked from commit 1e31c2c8)

(cherry picked from commit de0535d4)

(cherry picked from commit 05c46817)

(cherry picked from commit 6c6c671b)

(cherry picked from commit 6ed045c7)

(cherry picked from commit 51436a52)

(cherry picked from commit 55f72c0c)

(cherry picked from commit a85f6008)

(cherry picked from commit 191c8719)

(cherry picked from commit f0037bf2)

(cherry picked from commit e3597eaf)

(cherry picked from commit dfbad793)

(cherry picked from commit 1005075e)

(cherry picked from commit 76aed3dd)

(cherry picked from commit 4831fc1c)

(cherry picked from commit e9ee1b1a)

(cherry picked from commit 37d5a37b)

Job_id doesn't depend on retrials anymore when is written