zonefile: Verify mtime against full-precision timestamp
We've just used 1-second granularity mtime to check if a file has changed.
But if two updates happen within a calendar second, and knotd notices the first one and reloads the file, it might never notice the second change and continue serving the old file. We can see this happening in intermittent test suite failures in the debian continuous integration servers:
https://ci.debian.net/packages/k/knot/unstable/amd64Using nanosecond-granularity timestamps should make these problems go away.
Signed-off-by: Daniel Kahn Gillmor dkg@fifthhorseman.net