next: unsupported negative floating point number casts

[dastrm]

Using -float(1) anywhere within a field operator works, whereas float(-1) does not.

[tehrengruber]

This should be simple to add. The syntax type_name(scalar_val) is only implemented for literals, but float(-1) is actually a unary op on a literal, which is why it doesn't work. With GTIR the always somewhat artificial looking restriction that the argument has to be a literal is not required anymore anyway so we can fix this in one go.