spec: comparable explanation is self-contradictory

[rogpeppe]

In the spec, the documentation for comparison operators says:

In any comparison, the types of the two operands must unify or one of the operands must be null.

But that's not always the case in practice. For example, the following code is currently valid:

a: 1.0
b: int & 1
c: a == b

But this contradicts the spec, because a does not unify with b and neither operand is null.
For example this code:

a: 1.0
b: int & 1
c: a & b

gives this error:

c: conflicting values 1.0 and int (mismatched types float and int):
    ./x.cue:1:4
    ./x.cue:2:4
    ./x.cue:3:8

The spec should also state explicitly the rules for comparison against _|_, which clearly does not unify against anything except _|_ itself, but in fact appears to be comparable against anything in a similar fashion to null.

[mpvl]

We are considering changing the spec to allow any concrete value to be comparable.