Harden checkpoint and temporary-directory cleanup operations

[Widthdom]

Problem

Several cleanup and rollback paths manipulate checkpoint backups or recursively delete temporary directories after earlier validation. If a path changes between validation and use, or cleanup failures are swallowed, recovery behavior becomes hard to audit and can leave partial state behind.

Evidence

• src/CodeIndex/Cli/DbCommandRunner.cs: checkpoint restore delete/move/copy helpers around backup restoration.
• src/CodeIndex/Cli/DbCommandRunner.cs: temporary directory recursive cleanup after validation.
• src/CodeIndex/Cli/ProgramRunner.cs: upgrade script directory cleanup catches and ignores failures.

Requested improvement

• Revalidate cleanup targets immediately before recursive delete or restore moves.
• Make backup restore ordering/rollback auditable and fail predictably.
• Emit bounded diagnostics for cleanup failures instead of silent catches.

Duplicate check: no matching OPEN ISSUE found during the cdidx audit pass.

[Widthdom]

Work started on branch fix-issue3832-3812-3732-3810-3717-3737.