litewitness: witness should store full checkpoints

The signature a log makes on a checkpoint provides non-repudiation, i.e. recipients of such checkpoints can prove that the log issued that checkpoint. If a witness operator discovers log misbehavior (i.e. a split view signed by their witness) , they can use this signature to prove this misbehavior to others.

However, as it stands, litewitness only stores the root hash and size for each log, so while an operator could detect misbehavior, they cannot prove to others that they haven't made up this claim. Thus, litewitness should also either store the relevant signature made by the log or simply (like sunlight does) persist the whole checkpoint for each log.

Start by tracing how litewitness currently persists each log's root hash and size, then inspect how checkpoints and their signatures are represented. Done means a witness can provide verifiable evidence of the original signed checkpoint when reporting log misbehavior, rather than only the root hash and size.

Written by the indexing model from the issue text.