Disallow negative integers as indices

[71104]

Currently, int is allowed as index for indexed types (arrays and strings). The reason we can't restrict this to uint is that we want to be able to use expressions like - n 1 as indices. The current implementation of the - operator types to int even if both operands are uint because the result of a subtraction may be negative.

A possible solution is to implement proper 32- or 64-bit unsigned integers and wrap around when a negative value is produced, but this would raise the problem of deciding the width. A possible solution to this would in turn be to provide both 32-bit and 64-bit integer types.

[71104]

Another possible solution is to default to 32-bit width on 32-bit platform and 64-bit width on 64-bit platforms, but this would make it difficult handling 64-bit values on 32-bit platforms.

[71104]

Based on #77, wrapping integer values is no longer an option. Numeric types must now reflect the numeric sets used in math.

Another option would be to implement semi-circular arrays, like in Python; negative indices would refer to elements from the end (i.e. -1 refers to the last element, -2 to the second-last, etc.). This would justify integer indices.