afs-analyze-avoid-potential-loop-20030423

afs-analyze-avoid-potential-loop-20030423

avoid potentially looping forever in e.g. afs_FetchStatus because
we exit afs_Analyze too early to increment busyCount, and despite having
areq->busyCount non-zero we have servers still not_busy which are already down
that we'll never talk to again. that combined with no conn means we never
touch most of afs_Analyze, exiting early with shouldRetry set.

incrementing busyCount whenever we exit with shouldRetry set should
ensure we can never end up looping forever.