Deprecate `has(map.key) -> bool`

[hudlow]

has() is strange with respect to maps. It only works with string keys that are identifiers in the lexicon, and it suppresses what would otherwise be errors. (I could be wrong, but it doesn't seem like has(message.field) -> bool actually suppresses an error that its naked expression would result in.)

Plus, we already have key in map which seems much more idiomatic, eliminates error suppression of a subexpression, and can be used with any key value.

[TristonianJones]

Unfortunately, since has is a parse-time macro anything that looks like a field selection will be supported since the parse time won't know what the type of the select operand is. I agree that it not ideal though.