The db crash happens when calling `ezic_db:get_all(X)` when `X` is not
a valid table name. `ezic_db_ets` crashes, causing the `ezic_db`
server to crash, tossing the database with it. The solution prevents
ezic_db_ets:get_all/1 from crashing, and instead returns an error
value.
The simplification here eliminated some code through
deduplication. The `handle_call` functions could be improved
similarly.
The `ezic_db` server would crash on a bad lookup (no_zone,
ambiguous_zone, etc.). The old error notification method for bad
lookups (crashing) was killing the server and apparently eliminating
the in-memory tables.
