Retry unavailable cloud capacity

[ClaudiuSchuster]

Summary

• distinguish a completed Proton cloud-capacity failure from an in-progress refresh
• retry unavailable capacity reads after 30 seconds instead of leaving the card stale for a full successful-refresh interval
• retain the last capacity result across ordinary state polls and replace it after a successful retry
• make the independent cloud-capacity interval configurable as 5, 15, 30 or 60 minutes, defaulting to 5 minutes and applying saved changes immediately
• make the repository safety scan work in both normal checkouts and Git worktrees
• prepare the 0.8.3 patch release and matching Arch candidate references

Verification

• git diff --check
• make check
• isolated GTK Broadway widget checks and Preferences layout review
• Super-Linter 8.7.0
• Arch Linux smoke checks
• Arch package build, metadata inspection, installation and helper checks for 0.8.3-1

The live mount, VFS cache and remote data were not mutated during diagnosis or validation.