Resource manager: improve OPH_START_ERROR handling

[minomicetto]

Consider to retry the submission automatically when an OPH_START_ERROR (critical error) occurs (master: src/oph_rmanager.c: line 61).
The retry aims at re-submitting commands that have been rejected by the resource manager, maybe to its internal errors or low available resources.
A lot of rejects often occur in case of massive operations.
A new configuration line in etc/server.conf is advisable.

[minomicetto]

A preliminary support to automatically retry the tasks rejected by the resource manager has been added from release 0.10.5.
The procedure is based on two parameters of etc/server.conf: AUTO_RETRY and BASE_BACKOFF. The former is the maximum number of retries, the latter specified the maximum time interval between two consecutive retries (in seconds).
Setting AUTO_RETRY to a positive integer value enables the procedure.
Re-submission time is randomly chosen in interval [0, BASE_BACKOFF]. BASE_BACKOFF is doubled after each retry of the same task.