failpoints on disk reads

[spacejam]

trafficstars

[spacejam]

This needs to be performed in a different way than how our current failpoints are being handled, as when a failpoint is triggered, all writes from that point onward until the next fsync are considered to be stable. For this work, we need to relax the assumption, and only ensure that no operation ever panics. Errors on disk reads should always return an Err, rather than panicking, so this can actually be a much simpler model than the existing failpoint test.