Add support for initialization scripts

[taole33]

Fixes #3066

Description

This PR adds support for initialization scripts in MongoDBContainer.

When an initialization script is placed in /docker-entrypoint-initdb.d/, the official MongoDB image restarts the container process after executing the script. This behavior previously caused WaitStrategy timeouts because the container would stop and start again.

Changes

• Added withInitScript(String scriptPath) method to MongoDBContainer.
• The method copies the script to /docker-entrypoint-initdb.d/init.js.
• It automatically configures the WaitStrategy to expect two "waiting for connections" log messages, ensuring the container is fully ready after the restart.

Verification

• Added MongoDBInitScriptTest to verify that the container starts successfully with an init script.

[eddumelendez]

org.testcontainers.containers.MongoDBContainer is deprecated. can you moved to org.testcontainers.mongodb.MongoDBContainer?

[taole33]

I moved to org.testcontainers.mongodb.MongoDBContainer.
Also, For the tests, instead of creating a new file named MongoDBInitScriptTest.java, I added the test cases directly to MongoDBContainerTest.java.

[eddumelendez]

does this work with ReplicaSet and Sharded mode? New MongoDBContainer implementation has withReplicaSet() and withSharding() methods. It would be nice we can make sure this work with all implementations

[taole33]

I updated withInitScript to support ReplicaSet and Sharding modes.

In these modes, the default entrypoint mechanism conflicts with the cluster initialization scripts or is ignored. To fix this, I modified the logic to isolate the init script from the default directory when running in Cluster mode, and instead execute it manually after the container has started.

Reason

In Cluster modes (ReplicaSet/Sharding), relying on the default /docker-entrypoint-initdb.d/ mechanism causes issues:

• ReplicaSet: The default entrypoint triggers a restart, causing a race condition with rs.initiate().
• Sharding: The entrypoint is overridden, so the script is ignored.

Changes

• Added logic to detect Cluster mode.
• If Cluster mode is enabled, the script is mounted to a temporary location (isolating it from the default entrypoint) and executed manually after the cluster is fully initialized.
• This ensures data consistency and avoids startup conflicts.

Fixes #3066

[eddumelendez]

is this required?

[eddumelendez]

is this required?

[eddumelendez]

remove as

[eddumelendez]

Can we use plain english, please?

[eddumelendez]

let's remove it. There is no difference with the previous test

[eddumelendez]

let's remove it. There is no difference with the previous test

[taole33]

@eddumelendez
Thank you for the review. I have updated the implementation to address your feedback.

In this update, I also addressed an edge case regarding filenames containing special characters (e.g., # or spaces) which wasn't handled in the previous version.
Since standard resource loading can sometimes misinterpret these characters as URL fragments, I implemented the file transfer logic to explicitly handle raw bytes. This ensures robustness for any filename.

I believe this covers all cases. Let me know if you have any further questions.