Encoding of interpreter-level script representation

Sieve scripts might be loaded many times again and again while processing incoming messages. While it is possible to store `interp.Script` objects in memory, a practical implementation might want to avoid doing so as the size and amount of scripts can be significant. One potential solution is to define an efficient encoding of `interp.Script` data so it can be quickly restored from external cache on-demand without the cost of parsing the original Sieve code.