In theory, it is possible to have the same multi-layered bloom filter (MLBF) for different sets of inputs, because new additions could already be covered by existing bloom filters.

This situation does currently not happen in practice, because the backend uses a new salt (for its salted hashes) to generate the bloom filter.

But it would be nice to address it.

Emphasize that signedDate is a Date instance instead of a number.

And replace check for signedState with signedDate, because the latter
captures my intent (the original implementation checked signedState
because signedDate used to unconditionally return a Date instance,
even for invalid dates, but that was changed in bug 1583844).

https://hg.mozilla.org/mozilla-central/rev/05fa5a623f24
https://hg.mozilla.org/mozilla-central/rev/22b221cc4265