Initial action implemention

[omus]

Implements a working solution for capturing output from all jobs in a matrix. See https://github.com/orgs/community/discussions/26639 for general discussion on this GitHub issue.

See the design.md file for details on the implementation details and design decisions behind this action.

Depends on:

• Initial implementation job-context#1

[omus]

I noticed if you re-run a job you can end up finding more artifacts than the strategy.job-total. This detection scenario seems flawed in that multiple re-runs could result in us falsely believing all the jobs had completed when in fact we may have not successfully run one of the jobs. I've now dropped the artifact count logic.

https://github.com/beacon-biosignals/matrix-output/actions/runs/10376312018/job/28728003750#step:3:54

[omus]

Looks like I got a natural race condition on "Setup Complex" https://github.com/beacon-biosignals/matrix-output/actions/runs/10376653588/job/28729096768?pr=1 just after I added the artificial race condition test.

[omus]

Checkout is causing me grief now:

Error: Repository path '/home/runner/work/_actions/beacon-biosignals/matrix-output/466c693173d9d5b623e6125d66afb817477e5a83/job-context/repo' is not under '/home/runner/work/matrix-output/matrix-output'

– https://github.com/beacon-biosignals/matrix-output/actions/runs/10377452845/job/28731668322?pr=1

[omus]

Just fixed a bug where when two jobs wait for each other they weren't finding their sync artifacts due to comparing a numeric value with string in jq: https://github.com/beacon-biosignals/matrix-output/actions/runs/10414887856/job/28844541439?pr=1

I'm rather glad I hit this minor bug has it demonstrated that is possible for two jobs to have a complete set of outputs. Previously, this was just a theoretical problem.