Program functions can be declared as partial with "let/val partial".
Similarly to "diverges", partial code cannot be ghost, however it does not need to be
explicitly specified as partial.

Fixes #184.

Bench reestablished

The naming is now "consistent" in the sense that a transformation
argument is not taken in place of another.

However, it does not solve all the issues related to using
the IDE without the "introduce premises" enabled.

For first_order_matching, adding a bound_vars set to disallow substitution
by terms containing bounded vars.

- copy from a transformation to a proof node works, adding
  the transformation below the target node

- when copied transformation as less or more subgoals, then
  only the first subgoals are copied

- explicit error message in case of invalid copy

The feature is not yet fully implemented (e.g. escape characters).

- range types that fit in 31-bit signed integers are mapped to OCaml's type int
  (both literals and for loops)
- a for loop on a type that is mapped to 'int' in the OCaml driver is
  translated to an OCaml for loop

Since we are now testing for the inclusion of [val] declarations inside driver,
such a declaration was causing the test to fail. A possibility is to add a
driver to the test.

it returns (o,d), o true means there are obsolete goals, d means there
are missed objects (goals, transformations, theories or files)
that are now detached in the session

Note: there are several bugs remaining to solve, such that the fact that
missed transformations or lost instead of detached.

it now uses the module core/Pretty, that is generalized so as
to take ident_printer as arguments.
Notice the very nice use of first-class modules !

TODO: a bug remain when printing ident with space in them
TODO: remove the tables in printer_args

We need to discuss with Andrei about the use of "infix " in
infix identifiers which appears to be a problem for parsing
transformation arguments.
Anyway, we don't understand the specific hacks for "mixfix []"
and "mixfix [<-]" in Pretty.ml. Why not similar hacks for "mixfix [..]"
for example?

command "search" and transformations taking idents as arguments now can
support qualified idents and infix symbols.

For example, "search (+) (*)" returns the distributivity axioms

FIXME: "search Int.(+)" fails, probably missing namespaced for
imported modules