Better cache lock

[edsko]

Currently we create a directory hackage-security-lock in order to lock the cache; however, if the user interrupts the program (CTRL-C twice, for instance), we won't get a chance to remove this lock again and any subsequent attempts to check for updates will fail until the directory is manually removed.

A stop-gap measure would be to add a flag to ignore an existing lock, or ignore it automatically if the existing lock is older than a certain threshold; a proper solution would be to use some proper file locking but I don't know how to do that in a platform independent way.

[snoyberg]

This should be fixed by #203, with the caveat of #205. In other words, I think #205 supersedes this issue.

[dcoutts]

Agreed. Ta.