make submit --dry work with S3

[belforte]

looks like current dry run is based on TaskWorker/Actions/DryRunUploader preparing a tarball called dry-run-sandbox.tar.gz and uploading it to crabcache "by name" and then crab submit --dry finds it there (so if a user were to run submit --dry from two tasks at same time, there would be a mess).

Best way to do this in S3 seems to add another objectype and hanlde this like clientlog or twlog, placing it in the task directory of the user, since the taskname is known at the time that all those actions are done.

This requires another change to RESTCache, but then changes to client and DryRunUploader should be very small.

To be tested.

[belforte]

so I have three actions

• change DryRunUploader use S3 to upload tarball needed for DryRun #6465
• add runtimeSandbox handling to RESTCache add "runtimefiles" objecttype to RESTCache  #6552
• change CRABClient submit --dry to use runtimeSandbox from S3 make submit --dry capable to use runtimefiles from S3 as dry-run-sandbox.tar.gz CRABClient#4991

and possibly a fourth one

• avoid to duplicate user input sandbox inside runtimeSandbox - defer this to make preparelocal use S3 for tarball and unify with --dryrun #6544

[belforte]

all done. closing