The goal of this method is to hide some implementation details of FTP
server systems from the programmer. In particular, many FTP systems will
automatically disconnect a user after a relatively short idle time or after
a transfer is aborted. In this case, 'Net::FTP::AutoReconnect' will simply
reconnect, send the commands necessary to return your session to its
previous state, then resend the command. If that fails, it will return the
error.

It makes no effort to determine what sorts of errors are likely to succeed
when they're retried. Partly that's because it's hard to know; if you're
retreiving a file from an FTP site with several mirrors and the file is not
found, for example, maybe on the next try you'll connect to a different
server and find it. But mostly it's from laziness; if you have some good
ideas about how to determine when to retry and when not to bother, by all
means send patches.

This module contains an instance of 'Net::FTP', which it passes most method
calls along to.

These methods also record their state: 'alloc', 'ascii', 'authorize',
'binary', 'cdup', 'cwd', 'hash', 'login','restart', 'pasv', 'port'.
Directory changing commands execute a 'pwd' afterwards and store their new
working directory.

These methods are automatically retried: 'alloc', 'appe', 'append',
'ascii', 'binary', 'cdup', 'cwd', 'delete', 'dir', 'get', 'list', 'ls',
'mdtm', 'mkdir', 'nlst', 'pasv', 'port', 'put', 'put_unique', 'pwd',
'rename', 'retr', 'rmdir', 'size', 'stou', 'supported'.

These methods are tried just once: 'abort', 'authorize', 'hash', 'login',
'pasv_xfer', 'pasv_xfer_unique', 'pasv_wait', 'quit', 'restart', 'site',
'unique_name'. From 'Net::Cmd': 'code', 'message', 'ok', 'status'.
'restart' doesn't actually send any FTP commands (they're sent along with
the command they apply to), which is why it's not restarted.

Any other commands are unimplemented (or possibly misdocumented); if I
missed one you'd like, please send a patch.