Supports very simple programs that load integer immediates. It starts
the main, loads integer in registers, and return correctly.

Addition in Mach not yet supported, but should not be hard to add them.

Function calls are not yet supported.

The ABI for now is the same as the RiscV, with a small twist: $ra is
first loaded in a user register, then this user register is pushed
(instead of pushing $ra straight away).

ça va un peu plus loin!

Consider:
  struct P { int x, y; }
  struct S { struct P p; }

  struct P p0 = { 1,2 };
  struct S s1 = { .p = p0; .p.x = 3 };

ISO C99 and recent versions of Clang initialize s1.p.y to 2, i.e.
the initialization of s1.p.y to p0.y implied by ".p = p0" is kept,
even though the initialization of s1.p.x to p0.x is overwritten
by ".p.x = 3".

GCC, old versions of Clang, and previous versions of CompCert
initialize s1.p.y to the default value 0.  I.e. the initialization
".p = p0" is forgotten, leaving default values for the fields of .p
before ".p.x = 3" takes effect.

Implementing the proper ISO C99 semantics in CompCert is difficult,
owing to a mismatch between the intended semantics and the C.init
representation of initializers.

This commit turns the delicate case of reinitialization above
(re-initializing a member of a composite that has already been
initialized as a whole) into a compile-time error.

We will then see if the delicate case occurs in practice and needs
further attention.

The ISO C99 standard allows cast only if the type name involved
is either a void type or a scalar type.  
For compatibility with GCC and Clang we used to support casting
a struct or union to exactly the same struct or union type.
That does not seem useful in practice and complicates conformance
testing.  This commit gets rid of this exception to the C99 rule.

Bug 23310

Instead of overwriting the initializer of the anonymous member we
should just keep it.
Bug 23353