feat: reusable containers

[matthiasschaub]

adresses #109

Todo:

• [x] Write documentation about how to reuse containers.
• [x] Warn the user if with_reuse in use but ryuk is disabled.

Open questions:

• Should reuse_enable also be configurable via environment variable?
• Currently to make reuse work ryuk needs to be disabled. Should the user do this manually? -> if so warn the user if with_reuse in use but ryuk is disabled.
• Current tests are sensitive to users ~/.testcontainers.properties. This file should not be present during test run.
• Should each container class have the possibility to state if this container can be reused? (DockerContainer.reusable: bool = True)

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 93.33333% with 3 lines in your changes missing coverage. Please review.

Please upload report for BASE (main@0ce4fec). Learn more about missing BASE report.

Files	Patch %	Lines
core/testcontainers/core/config.py	80.00%	2 Missing :warning:
core/testcontainers/core/container.py	96.55%	0 Missing and 1 partial :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #636 +/- ## ======================================= Coverage ? 77.27% ======================================= Files ? 11 Lines ? 616 Branches ? 93 ======================================= Hits ? 476 Misses ? 113 Partials ? 27 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[alexanderankin]

will review and merge if you have no other plans for changes

[matthiasschaub]

@alexanderankin we would welcome a review from you, thanks. They are no further plans right now except addressing what ever comes up in the code review.

[kiview]

Thanks for looking into this @matthiasschaub 👋 We have no clean spec for this features in Java / across languages, so for now I would suggest, we mostly mirror the Java implementation, including its limitations. We can sync on a cross language spec in the future that provides a better DX, but for now I would strongly favor an implementation that mirrors Java.

I already left some comments within the PR.

Currently to make reuse work ryuk needs to be disabled. Should the user do this manually?

Given the above, if a container has reusable set, it must not be registered with the Ryuk cleanup label (see tc-java).

[matthiasschaub]

Thanks for looking into this @matthiasschaub 👋 We have no clean spec for this features in Java / across languages, so for now I would suggest, we mostly mirror the Java implementation, including its limitations. We can sync on a cross language spec in the future that provides a better DX, but for now I would strongly favor an implementation that mirrors Java.

I already left some comments within the PR.

Currently to make reuse work ryuk needs to be disabled. Should the user do this manually?

Given the above, if a container has reusable set, it must not be registered with the Ryuk cleanup label (see tc-java).

Thanks for the review @kiview! I agree. It is sensible to follow the Java implementation.

In commit 1ea9ed16a0dd73e11d762808aed5655d44e270be I do not create a Reaper instance during container start-up if reuse is enabled and container has been started with with_reuse. This works as expected as long as no other container is started without with_reuse. In that case a Reaper instances is created which will remove the reuse container as well. Is there a way to explicitly exclude a container from the Reapers routine?