Improve storage fail detection by putting the object

[ne2pit]

Sometimes storage might be alive but work very slow or go to read-only mode. Current fail detector uses ListObject method to check it's liveness. Such check is unreliable.

This PR changes check method to PutObject. Data is pseudo random generated. The object size and the timeout is configurable. Default values are 1MiB and 30s. The object has predefined name always. So there is no additional storage space waste on sequential checks.

[usernamedt]

LGTM, let's wait for the context PR to be merged. And the last thing - add a note to the docs about the new setting.

[ne2pit]

Returned ListFolder check as we discussed. And now our assumption to mix it with context PR become different. Adding ListFolderWithContext is trivial for most of storages except SSH. It is impossible to make it with current SFTP library on the first sight. What do you think? Should we still wait for context PR merge?

[usernamedt]

Returned ListFolder check as we discussed. And now our assumption to mix it with context PR become different. Adding ListFolderWithContext is trivial for most of storages except SSH. It is impossible to make it with current SFTP library on the first sight. What do you think? Should we still wait for context PR merge?

We still have the WALG_FAILOVER_STORAGES_CHECK_TIMEOUT so I guess it is ok to simply call the ListObjects(). Since it is a simple API call and we don't upload anything in it, it is unlikely that it would hang indefinetly.