Lockfile handling doesn't account for an already-terminated pid

[lycurgus]

See urbit/urbit#6795 - when u3_disk_acquire attempts to kill the pid it finds in the lockfile, it doesn't account for the pid already being gone. In such a case kill will return -1 with ESRCH and vere can (presumably) delete the lockfile and proceed. Currently the operator has to investigate, determine that the lockfile's pid is already gone, and delete the lockfile manually in order to boot.

It seems like disk.c needs another check that handles ESRCH coming back from kill. Potentially just amending line 780 to read if ( -1 != ret || errno != ESRCH ) { - but that seems somewhat like a loss of legibility. I figured I'd raise an issue since my u3 isn't really at a level to raise a PR.