[RFC] Wrapped negation of unsigned integers

[HertzDevil]

Regular negation of unsigned integers always leads to an underflow for anything other than 0, but wrapped negation is still useful when dealing with binary data or low-level code. Crystal however doesn't support them:

x = 1
-x # => -1

x = 1_u32
-x # Error: wrong number of arguments for 'UInt32#-' (given 0, expected 1)
&-x # Error: wrong number of arguments for 'UInt32#&-' (given 0, expected 1)

This means one is left with these less clear alternatives:

~x &+ 1 # => 4294967295
x.class.zero &- x # => 4294967295

Thus I propose that the unsigned integer types (UInt8, UInt16, UInt32, UInt64, UInt128) define the argument-less #&- method, which performs wrapped negation. There are some possible extensions:

• Signed integer types could define #&- too, but the only affected value is the minimum representable integer of each respective type (in contrast to a hypothetical #- working on exactly one value from each unsigned type), so that's not as useful.
• The argument-less #&+ is exactly the same as #+ for integers, so that's probably useless as well.