Fill NAs with 0s by default in prepare_data.incidence2

[pratikunterwegs]

If a default value of FALSE is more likely to cause issues, why not set it to TRUE and throw a message informing the user about the operation? In fact, I would argue that this not be an argument at all but rather operate under a value of TRUE assumption, which should be documented as part of the data argument and discussed in the details section of the function.

Originally posted by @jamesmbaazam in #78 (comment)