Hex literals can be invalid UTF8

[adesutherland]

The compiler accepts hex literals 'FFFFFF'x and these are store as utf8 strings.

Unfortunately if the hex literal is not valid utf8 then later the character rxas instructions fail.

Actions

• The compiler needs to validate strings are well formed utf8
• Level b needs to support a binary datatype
• Level c will need an approach to distinguish (transparently to the user) between binary and utf8 data

[Peter-Jacob]

Introducing a dedicated binary type feels like overkill, given how limited its actual use case appears to be.

If we restrict the UTF-8 string to just the characters 0–9 and a–f, then each character is a single-byte UTF-8 code point (i.e., standard ASCII). We could define a new instruction that converts such a string into a plain char array. If the string contains invalid characters, we could raise an error — potentially even at compile time.
From my perspective, adding a full binary data type might introduce more complexity than we need at this stage.

[rvjansen]

I am not against a BYTE type, which I am using regularly in NetRexx (Java's byte type, that is)