In the -2c mode, in presence of extern types, I was defining them as int.
The idea was to have something that compiles, and then to ask the user
to redefine its extern types.

Now I define extern types as string (char*).

The rationale of this change is that before, the generated code was only
working for int, which is pretty useless as an extern type.

Have string by default is actually much more useful, for instance to use
string in a Lustre program.

Also, adapt to changes done in lutils, where the string type has ben added.
(sha:8619aa285ca8322ab9, i.e., > lutils  1.49.10)

(It would have avoided issue #8)

close #8

Remove a lot of warnings (considered as errors by dune).

cd rdbg #sha 4a54c7a1181e1200670f3eaf14af1be32823cd04

Indeed the API of Format.formatter_out_functions changed in 4.06.0

To turn around this problem, I've used pp_get_formatter_out_functions
and extend it result

Rationale: a step to make time traveling work correctly.

It causes bad error messages in the case were the parametric node is undefined.

e.g., the equation

  S=if(M[n-1]) then (entiernat<<n>>(M)-puiss<<n+1>>(2)) else entiernat<<n>>(M);

raises "Error: Bad number of (static) arguments: 0 expected, and 1 provided."
when entiernat is undefined

It seems useless because no test is broken. The comment mention node
parameters that themselves have parameters. But it is the case for
test/should_work/Gyroscope2.lus, and it is still working.

update the Arduino demo

nb: i should really add 'make test' in the git commit hooks!

Remove unused files along the way (myGenlex.ml/mli)

By not catching errors, rdbg and (thus) lurette were silently do nothing
in presence of errors in Lustre files.

The idea is to be able to take into account clock in node profiles
when generating the main simulation loop, and to avoid to read inputs
that are not present.

The generated code for tests on boolean output was wrong. Indeed, it
was generating

  if (o1 == _true) {

instead of

  if (*o1 == _true) {

nb : a warning was emited