Amas :
 * Rework the way to add agents to give more option to user, can now choose multiple agents and choose agent destination

From : add_agent(self, experience_name: str, args: List = None)
To : add_agent(self, experience_name: str, client_ip: str = None, agent_name: str = "agent.py", args: List = None)

param experience_name: name of the experience folder
param client_ip: if the agent should be created in a specific device, you can specify an ip address, otherwise the Amas will try to share the work between the devices
param agent_name: if using multiple kind of agent, you can specify the relative path in the experiment directory to the agent file to use
param args: if any argument is needed to initiate the new agent

Added Mail class, that can be used to communicate between agents, a mail contain the sender id, the cycle of the mail creation as well as a payload that can be anything serializable.

Added CommunicatingAgent class, that can be used instead of agent
 * A mailbox that work like a stack of mail
 * A method to send mails
 * And a method de receaive them

Scheduler :
 * Fixed a bug where the mqtt client would lock himself making communication with the scheduler impossible