ci: send Slack notifications on CI failures

[anakin87]

Related Issues

• part of https://github.com/deepset-ai/haystack-private/issues/256

Proposed Changes:

Instead of sending DD events for CI failures and then Slack notifications from DD monitors, skip DD

• uses the composite action https://github.com/deepset-ai/notify-slack-action
• adopt the composite action in all integrations workflows
  • to make sure we get a single notification when a matrix workflow fails (not one for each element in the matrix), I created a new lightweight job (uses ubuntu-slim)
• updated the template for new integrations to use this new pattern
• removed the old send_failure composite action (integrating with DD)

To make this work, before opening this PR, I completed the setup described in the Slack documentation.

How did you test it?

In this PR, I temporarily made Anthropic workflow fail and got the expected notification.

Since this was a test, I avoided tagging people. The actual action would tag @here in the Slack channel.

Checklist

• I have read the contributors guidelines and the code of conduct
• I have updated the related issue with new insights and changes
• I added unit tests and updated the docstrings
• I've used one of the conventional commit types for my PR title: fix:, feat:, build:, chore:, ci:, docs:, style:, refactor:, perf:, test:.

[anakin87]

Nvidia tests failing due to #2838 (unrelated to this PR)

[bogdankostic]

Looking good in principle, just wondering why we use a different pattern in CI_license_compliance.yml.

[bogdankostic]

In all other workflow files we have the notification step as a separate job.

[anakin87]

Yes, in other workflows, we are using a separate job for this reason

to make sure we get a single notification when a matrix workflow fails (not one for each element in the matrix), I created a new lightweight job (uses ubuntu-slim)

In this case, we don't need a matrix: the license check only needs to run on a single operating system and with one Python version. Where possible, it's better to have a step instead of a job because GitHub does not need to spin up a new container.