default actions in case explicit options are missing.

The command line has the following structure:
    
    why [options] [[file|-] [-t <theory> [-g <goal>]...]...]...

Summary:
    
    Qualified theories come from the library, not from file.
    When the driver is not specified, pretty-print theories.
    When neither --prove nor --output are given, print tasks.
    When no theory is specified for a file, use every theory.
    When no goal is specified for a theory, use every goal.

Examples:

    % why.opt
    [shows usage information]

    % why.opt test.why
    [prints the contents of test.why]

    % why.opt -t int.Int
    [prints the library theory int.Int]

    % why.opt -D drivers/z3.drv test.why
    [sends every task from test.why to stdout]

    % why.opt -D drivers/z3.drv -o directory test.why
    [creates ./directory/ if it's does not exist and
     sends every task to a separate file in ./directory/]

    % why.opt -D drivers/z3.drv --prove test.why -t ThA
    [calls the prover for every goal from theory ThA in test.why]

    % why.opt -D drivers/z3.drv --prove test.why -t ThA -g G1 -g G2
    [calls the prover for G1 and G2 from theory ThA in test.why]

    % why.opt -D drivers/z3.drv -t int.Abs -g G1 test.why -t ThA
    [prints G1 from int.Abs and every goal from ThA in test.why]

It's not more expressive but much nicer than epsilon.

- provide several helper functions in Task
- Pretty.print_named_task will also print Use and Clone decls

This will allow transformations to apply use and clone.
Also, this will allow to simplify the driver structure.
Memoisation modulo env/clone/use requires some further 
adaptation (but this would be needed anyway).

  as it is most often ignored
- check for empty map in t_subst/f_subst
- bugfix: don't forget nested match statements in Decl and Compile_match

  (TODO gestion des commentaires, des match, forall, ...)
   - call_provers : Utilisation de Unix.create_process au lieu de Sys.command

  move this implementation to eliminate_let.
- initial commit of eliminate_definition : the goal is
  to translate definitions by match into series of axioms.

- check pattern well-formedness in Pattern
- add arguments to exceptions in Ty and Term

(unusable so far, must eliminate match statements, too)