Fixes #16

nb: does not really work yet.

Update: add the previous state in argument of the fault function.

New: add a fault functions to the Algo (to inject fault when a legitimate configuratiuon is reached).

nb: I needed to  split Main to avoid a Module self-loop  (as I need to
use do_step in Daemon).

that is now genarated all the times by rdbg

The  rationale is  that the  Algo.spid that  gave processes  access to
their pid via their neighbors was a bit weird. Also, it makes the anonymity
of the algorithm more explicit: an access to the pid could appear anywhere
in the program, while now, the pid must be in state.t!

Now, if the  user needs the pid,  he must store it  explicitly in the
state via the initialization function.

To force that, I've changed the Register interface (and thus GenRegister)

The rationale is that:
- in lurette mode, they should be the same for all algos already
- in batch mode, action declarations are ignored anyway (except in custom mode)

also, use (back) the odoc make rule now the image has been updated

Indeed, during the explanations, it makes more sense to start with the state.

Doc: rework the documentation of the Algo API.

Its basename is the basename of the name of the .dot file.

Also provide a =--gen-register= option generates such a registring ml
file. --gen-register also generate a state.ml skeleton file.

The Makefile test/Makefile.inc also suppose that the types of the state
is defined in state.ml

To do that, I've reorganized the code

Test: with ocaml 4.08, it is impossible to load the same function twice.
Hence the need to define p.mli files for graphs with several nodes.

Also, track missing files and the alea-coloring exemples.

Add use it to add an oracle to the BFS

All the get_* functions (that are of no interest for users) are now in
the Register module.

Things that are currently broken:
- the ability to init vars in the dot
- controling the name of local vars in the rif output

The idea is to use the Obj module (arg !!!), which is safe thanks to
the fact that there is now only one register function, which forces all the
'v to be bound to the same type.

Indeed, by using the Env.nset and Env.nget defined in
0d6575a7 a pb that was hided by the
use of Env.set and Env.get was triggered.

The bug was that variables initialisations were done several times.

In sasa.ml, the way I was updating the Env.t by iterating on the
variables was forcing to "close" expressions that were calling init
functions.

The rationale is that
 - it is not always the cmxs that is necessary .e.g., rdbg uses .cma or .ml
 - it makes more sense to refer to the source code