fix(git): check destination emptiness without full readdir

[AdeshDeshmukh]

Avoid reading an entire directory just to determine if clone destination is empty. Use ReadDir(1) to detect emptiness efficiently while preserving error behavior for non-empty and inaccessible paths.

Description

This PR improves clone destination validation in pkg/lib/external/git/clone.go.

Previously, checkDestination used os.ReadDir(path) to determine whether a directory is empty, which reads the entire directory contents even though only an empty/non-empty check is needed.

This change replaces that with a one-entry probe:

• os.Open(path)
• ReadDir(1)

Behavior is preserved:

• err == nil => directory is non-empty
• err == io.EOF => directory is empty
• other errors => return wrapped inspection error

Validation:

• go test ./pkg/lib/external/git (pass)

Linked issues

N/A

[gorkem]

Can you add the sign-off messages to your commits. Click on the failing DCO check above for how to instructions.

[Copilot]

Pull request overview

Improves clone destination validation in the Git external helper by checking whether the destination directory is empty without reading the entire directory listing, reducing unnecessary I/O for large directories.

Changes:

• Replace os.ReadDir(path) full directory reads with os.Open(path) + ReadDir(1) to detect emptiness.
• Add io import to handle io.EOF from the one-entry probe and preserve existing error semantics.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[gorkem]

@AdeshDeshmukh please fix failing DCO check before we can review this.

[amisevsk]

Please reduce this changeset to only address the issue mentioned in the PR title; changing annotations used by Kit and adding legacy options should be done in a separate PR (personally I don't think this is very high priority at the moment).

Also, there are merge conflicts that need to be addressed.

[amisevsk]

This change, while generally something we probably should do, is unrelated to the purpose of this PR and should be broken out into a separate PR.

Additionally, it adds a decent amount of complexity for handling annotations; I think we're fine sticking with the ml prefix for now.

[AdeshDeshmukh]

@amisevsk Thank you for your thorough review and valuable suggestions. I've removed the annotation migration changes as you recommended — the PR now focuses solely on the clone.go directory emptiness check. I've also resolved the merge conflicts and ensured DCO compliance. Would appreciate a re-review when you a bandwidth. PTAL