Radical.pilot

- fixed pilot staging for input directories
- clean up configs
- disabled `os.setsid` in `Popen` executor/spawner (in
`subprocess.Popen`)
- refreshed module list for Summit
- return virtenv setup parameters
- Support for :py:mod:`radical.pilot.X` links. (eirrgang)
- use local virtual env (either venv or conda) for Summit


--------------------------------------------------------------------------------

- adapt flux integration to changes in flux event model
- fix a merge problem on flux termination handling


--------------------------------------------------------------------------------

- artifact upload for RA integration test
- encapsulate kwargs handling for Session.close().
- ensure state updates
- fail tasks which can never be scheduled
- fixed jsrun resource_set_file to use `cpu_index_using: logical`
- separate cpu/gpu utilization
- fix error handling in data stager
- use methods from the new module `host` within RU (\>=1.6.7)


--------------------------------------------------------------------------------

- added flags to keep `prun` aware of gpus (PRTE2 LM)
- add service node support
- Bridges mpiexec confing fix
- task level profiling now python independent
- executor errors should not affect task bulks
- revive ibrun support, include layout support
- MPI standard prescribes -H, not -host
- remove pilot staging area
- reduce profiling verbosity
- restore original env before task execution
- scattered repex staging fixes
- slurm env fixes
- updated documentation for `PilotDescription` and `TaskDescription`


--------------------------------------------------------------------------------

- added flag `exclusive` for tags (in task description, default
`False`)
- Adding Bridges2 and Comet
- always specifu GPU number on srun
- apply RP+\* env vars to raptor tasks
- avoid a termination race
- Summit LFS config and JSRUN integration tests
- gh workflows and badges
- ensure that RU lock names are unique
- fixed env creation command and updated env setup check processes
- fixed launch command for PRTE2 LM
- fix missing event updates
- fix ve isolation for prep_env
- keep track of tagged nodes (no nodes overlapping between different
tags)
- ensure conda activate works
- allow output staging on failed tasks
- python 2 -\> 3 fix for shebangs
- remove support for add_resource_config
- Stampede2 migrates to work2 filesystem
- update setup module (use `python3`)


--------------------------------------------------------------------------------

- fix uid assignment for managers


--------------------------------------------------------------------------------