badger backup system

[requilence]

for spacestore

• run incremental backup every minute
• every 1000 backups(empty skipped) do the full backup and remove all others after
• do the os.Sync before renaming backups from *.tmp
• in case it was not able to open badger on start recover from backup
• only in case backup restore was successful the old corrupted dir got renamed and the restored one take it place
• because we never rewrite most of the keys' values in spacestore(only object heads), incremental backup works effective and doesn't consume a lot of space

for localstore

• in case of error on open just reinit the db without backup
• corrupted repo left for debugging

Also Need to switch to this PR's branch https://github.com/dgraph-io/badger/pull/2033 The reason is that in the most cases the panic happens in the goroutine and it not possible to catch it in from the badger.Open caller

I think backuper may be used only on WIndows. Let's see how often this corruptions happen on UNIX

[github-actions[bot]]

New Coverage	43.7% of statements
Patch Coverage	47.1% of changed statements (105/223)

---

^{Coverage provided by https://github.com/seriousben/go-patch-cover-action}