Removes debug flag: debug_auto_model.
Some changes in counterexamples triggered by:
- (non counterexamples) transformations which have a specific case for
   model_trace but not for the new detection: this is intended as
   simplifications that would be done are often simplifications we want
   for counterexamples,
- Some locations are missing in variables introduced by SP/WP which should
  explain the rest.

This also disables projections for record in intro_projection_counterexmp.

Correct subst_filter to be consistent with new counterexample modification

We now use the information inside attributes to correctly associate a
label to each counterexamples. The information is carried inside
attributes which are preserved/collected during ce transformations and
printing to smt2.
The collection filled during the printing of the task is reused during the
printing of counterexamples to add "at label" where needed.

This makes Timeout and Unknown equivalent and remove the diff on floats
that differs on moloch and CI.

This updates values of already failing counterexamples which were changed
by the fix for z3/encoding_twin in 4b79dcf1.

This fixes a problem in the wp generation and eval match where it was
possible to create new variables with same labels (including model_trace)
which does not have the same type. This results in bad typing for
counterexamples.
In particular, when only one region of a type is mutable we can project
directly during wp (now we do the same but with the corresponding
model_trace).

model_trace from the model_context when defining a new variable

This way, (labels|tags|attributes) can be retrieved on counterexamples
names.

Fix example incremental.

- no option -get-ce and option in IDE anymore

  instead counterexamples are generated using prover alternatives

- for counterexamples, smt printer always prints incrementally: first the goal,
  then the ground hypotheses, then the others