Consider using PRECONDITION etc. in expr2c

[tautschnig]

As briefly discussed in #1726, expr2c does not enforce the expected structure of expressions. It may be worth revisiting this, and this should be done in one go instead of piecemeal.

[martin-cs]

If it is using things that should always be true for expressions, may I suggest DATA_INVARIANT is the way to go so eventually we can check these are enforced within the expr objects.