Better control for numerical limits

[cblc]

I think two features are very desirable:

1. Detecting overflow and underflow, issuing errors or warnings (it could be configurable if you want a warning or an error) when an overflow or underflow occurs. Sinclair BASIC issued an error when a number was "too big". Some of the open issues are actually side effects of allowing overflow/underflow. If you don't allow overflow/underflow to happen, you can close some issues.

2. A configuring macro that forces all numbers to be floating point. For some use cases where you want maximum simplicity, I feel like you might want to say "every number in the program is floating point". I find myself in this situation. IIRC, all numbers in Sinclair BASIC were floating point, although they were converted to integer when a function really needed a integer input.