Typecast constants in compile time

[neilsf]

Members of an expression are converted to the same type in runtime only. For example, the expression

DIM x AS FLOAT, y AS FLOAT
x = y + 1

will be compiled to the following runtime code sequence:

1. Push y as float onto stack
2. Push 1 as byte onto stack
3. Convert top byte on stack to float
4. Add top two floats on stack
5. Pull float result off of stack

We should avoid conversion in runtime and do the work in compile time.

Note: using floating point literals solves the problem, but programmers aren't always aware of this, e. g:

x = y + 1.0