feat: add simple object storage cleaner script

[Rakanhf]

Currently QFieldCloud keeps all project files using the non-legacy storage even if they are deleted by the user. This is due to enabled versioning on Exoscale and the lack of Exoscale mechanism to delete files after certain amount of time after they are deleted. This causes excessive file storage and therefore costs.

• Add purge_deleted_objects.py for scanning and deleting logically deleted objects in S3 compatable storages
• Add test.py for testing the script
• Add a README.MD

---

A small CLI to analyze and clean logically deleted objects in versioned S3-compatible buckets.

This script is provider-agnostic works with any S3-compatible service that supports versioning such as MinIO, etc.

It helps you to the following:

1. Detect whether a bucket has versioning configured.
2. Find keys whose latest version is a delete marker (logically deleted).
3. Report how much storage their historical versions consume.
4. Deleting all versions for those keys to free space.

Requirements

• boto3
• mypy-boto3-s3 Optional for type hinting

Usage

1. Scanning for logically deleted objects

python purge_deleted_objects.py my-versioned-bucket-name --retention-period <time-interval> --dry-run

2. Deleting logically deleted objects

python purge_deleted_objects.py my-versioned-bucket-name --retention-period <time-interval>

Options

Option	Description
bucket_name	The name of the target S3 bucket (Required).
--dry-run	Only scan and report statistics without deleting.
--force	Skip confirmation prompt (useful for automation).
--prefix <path>	Limit scan to a specific prefix (folder).
--retention-period <time-interval>	Only process objects deleted BEFORE this duration (e.g., 7 days, 30 minutes, 1 hour). Recent deletions are preserved.
--log-level	Log level (DEBUG, INFO, WARNING, ERROR, CRITICAL).

[duke-nyuki]

Task linked: QF-6039 [filestorage] Write a script that actually deletes the files from the object storage

[suricactus]

Did a quick round and this is very close to the final result.

The only major thing is that the date anchor is inverted, the rest are style, readability and maintainability comments.

[suricactus]

Is there a specific reason we use pytest for the tests? Not that I am totally against, but anywhere else we used the unittest module (or inherited with the django test framework).

Can you add as part of this PR a CI that runs this test every time we push, so we can guarantee the tests are working?

[suricactus]

Finally managed to have a look on tests. They are good, IMO they can become even better, check the comments

Pretty sure this is the last round of review from my side, we are almost there. Once addressed the review, take it as approved from my side.

The biggest requested change is to make the --retention-period a required parameter. The rest are typing and "don't shoot my foot" suggestion.

Since this is a critical piece of software, please request a review from some of the other maintainers of QFieldCloud.

[suricactus]

Finally we are here.

Would be great to extensively test it in real world scenarios.

@gounux can you please add an additional review before this gets merged?

[gounux]

Could test it on my local minio, seems to work alright.

Please find my review round here, partly focused on doc, usage and some small adjustments.

Thank you for this crafted script @Rakanhf :)

[Rakanhf]

I believe that now V1 of this script is ready @suricactus @gounux

Before I merge I will clean up the comments so it merges cleanly

[suricactus]

@gounux would you mind approving this PR and having it merged?