schema patch_of_a_patch: Type system in PostgreSQL is really weird, and doesn't allow even explicit cast from DOUBLE PRECISION to NUMERIC

schema patch_of_a_patch: to stop confusion caused by automatic naming of FKeys in MySQL, start giving them names (rest to be fixed in the next official patch)

schema patch_of_a_patch: reconstruct the role table's contents to the best of our ability from corresponding log messages (in MySQL)

api_change: beekeeper passes only the submit_log_dir to the Meadow, which may then decide how to capture submission logs best (thanks, Lel!)

api_change: Runnable/Process no longer caches a reference to Analysis. It can be had via $runnable_object->input_job->analysis if needed.

cosmetic/bugfix: refer to load_resource_usage.pl in usage legend instead of lsf_report.pl that is no longer around

bugfix: when using more than one standalone worker make sure their /tmp directories do not clash

The worker_temp_directory for standalone workers was set to
/tmp/worker_${USER}.standalone. If two standalone workers were running
on the machine, they could overwrite each other’s files.
The fix include the process ID ($$ in Perl) in the directory name.

skip undef values (do not even mention them) when performing store via insert (and let SQL layer initialize undefined fields)

schema_patch_of_a_patch: no need to enforce certain seconds to be INT (a particular limitation of LSF)

reverted to the old behaviour: running beekeeper.pl -logic_name will also set workers' resource_class

bugfix [to ENSCOMPARASW-136] : life_cycle() now performs autoflow for jobs that choose to exit via incomplete(0)-and-die

generalize the resource_usage loading script to work with those Meadows that support this functionality

moving code that assembles the bacct command line given an interval into Meadow::LSF::get_report_entries_for_time_interval and calling it from there

Also removed direct way of setting start_date and end_date and provided a generalized -source parameter for parsing from files/pipes