Overflow detection in simulation

[youben11]

No description provided.

[BourgerieQuentin]

As well should have tests

[umut-sahin]

Also, we need to check for errors after table lookups as they can result in overflows (imagine, [1, 2, 3000, 4] table and 2 is not observed in evaluation, but happened in simulation)

[BourgerieQuentin]

On my previous review I totally forgot the signed integer case, actually the encoding differs from unsigned and signed integers. On unsigned integers we let the padding bit clean, but on signed integers we use the padding bit as the traditional sign bit to let the leveled signed operations works natively then we rescale before tlu.
So we should check overflow in different way on signed (against 64bits) and unsigned (against 63bits). One of the issue is that the signed information is not available at TFHE level. Let's live discuss of that.