This modification impacts all parts of Batsim in which machines were involved.
The protocol has also been updated to allow the scheduler to send machine ranges
in its allocations. The previous syntax is still supported. Now, each comma-separated
value can either be 1 machineID or a range of machineIDs (syntax: machineIDa-machineIDb
to represent the closed interval [machineIDa, machineIDb]).

Please note that Batsim has been only been tested with old schedulers (that do not
support the new protocol syntax).

Separation of the network and the server in different files (which simplified the use of XBT_LOG). Minor fixes: missing enumerated cases in string conversion, use of a switch to generate a warning on future enumerated value misses. Minor cosmetic modifications.

If the energy is enabled (by setting the -p run option), the platform is assumed to contain energy information.
Such information is read by SimGrid itself but another ones, such as the 'sleep_pstates' property, are read by Batsim.
A lot of assertions has been put in the reading of the platform file which will help generating energy platforms.
An example of such a platform can be found in platforms/energy_platform.xml. This platform follows the same rules
that those of SimGrid energy platforms but assures pstate coherency via the 'sleep_pstates' property: it says
which pstates are sleep pstates and which pstates are its associated virtual pstates (for switching ON and OFF
each pstate). More information can be found in the file 'platforms/energy_platform.xml'

The scheduler can now send messages of the form TIMESTAMP:P:M=S where M is a machineID and S the asked new pstate
for machineID. Batsim does the stuff to simulate the time & energy consumptions of the switchON/OFF (if any, you
can also switch from one computation state to another). Once the pstate of a machine has changed, Batsim sends
it on the socket with messages of the form TIMESTAMP:p:M=S where M is a machineID and S the new pstate of the
corresponding machine (note the lowercase 'p' instead of an uppercase 'P').

To simulate the energy & time costs of the switch ON/OFF, the exact same mechanism as the one presented in
SimGrid energy 1O1 is used: computing 1 flop on virtual pstates whose computational power and power
consumption met our time and energy consumptions needs.

CMakeLists has been updated. Added some files which were missing in previous commits. Checked workload validity. Lots of TODO have been done: updating the export.hpp documentation, better displays...

The compile flag -std=c11 is now disabled (caused compile-time errors when using SMPI on my machine)

Some machine metadata are now stored. This is done to prepare the energy integration (in order to change a machine state, such a state must exist somewhere...). This will also allow to improve the Pajé trace generation in the future.

Modification des structures de données utilisées pour accéder aux jobs, ajout d'une méthode jobFromJobID. Le vrai ID du job est utilisé partout dans le code pour cohérence. Correction de divers warnings clang. Ajout de copies mémoire après les lectures depuis le fichier JSON en prévision d'un clean de la mémoire associée au fichier JSON. Ajout de vérifications concernant les données venant de la socket (si le job existe, si le job est dans le bon état, si les machines sont valides...)

La schedule résultante du simulateur est générée sous la forme d'une trace Pajé schedule.trace (visualisable via ViTE par exemple).

Modifications du CMakeLists afin de compiler en C11 (ou au moins en C99), pour éviter les vieilles contraintes de C89.
Ajout de l'include <sys/types.h> pour palier à un problème de compilation de /usr/include/xbt/str.h en C11.