Radical.pilot

- add tutorial for app level scheduling
- distutils are deprecated (bootstrapper)
- extended functionality for `startup_timeout` attribute in TaskDescription
- fix install of internals.rst
- fix option check in mpiexec
- implement startup signal
- make it simpler to assign session IDs to experiments
- make sure task tmpdirs exist
- overwrite `NumaNode.as_dict()`
- random changes to update old code snippets - please ignore in review
- simplify NUMA scheduling, improve examples


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

- allow oversubscribe on localhost
- close control subscriber on `close()`
- control PSI/J `exclusive` parameter for the LRM submission
- don't use a gpu in examples - that limits usable systems
- ensure scheduler sub-process terminates
- fix cancellation procedure (executor modules)
- fix rankfile / hostfile checks
- fix `task['proc']` handling (in `Popen._check_running()`)
- update py-versions for `ubuntu-latest`
- update references to RP classes
- update tutorials and docs


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

- `glob` does not understand POSIX globs :-(
- add `DOWNLOAD` staging action
- add resource configuration for `epcc.archer2`
- improve process watching
- ensure cancel arrives at executor
- external service support
- fix delta config
- fix example
- fix io redirection for service tasks
- generously increase proxy timeout
- remove MacOS support
- support non-ve mode
- only service rank 0 provides info string
- run examples in PR CI
- safer srun termination
- set default `scattered=False`
- support multi-rank services
- support ranks per node
- target state is interpreted *after* output staging
- track output files
- updated action versions
- use python to fetch URLs
- we expect task metadata to be a dict - enforce and document


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

- support numa aware app level scheduling
- apply cb lock to pilot updates
- fix task cancellation
- initial implementation of task priorities
- remove docker folder and link the tutorial repo in the README
- rename artifacts
- run example ci tests separately


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

- always log full version also
- ensure state setting
- updated docs and configs for Delta (NCSA)


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

- support for application level scheduling
- apply LM blacklist
- avoid uneccessary repr implementations
- ensure cancellation request is forwarded to agent and scheduler
- ensure full task info dict


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