Implement save/load

[Andrei486]

First step: implement saving and loading a command list to/from a file. Ensure that all commands that need to be added to the list are actually added, both on client and host side, and ensure that loading correctly re-creates the command list and can apply them on startup. (PR #98)

Second step: when the command list is being saved, "compress" the saved version, the current one stays unmodified. This process should keep only the minimum number of commands needed to re-create the diagram without any loss.

[Andrei486]

Idea for compressing a command list:

• Immediately remove all commands related to only element IDs that do not exist anymore in the diagram.
• Go for each remaining ID, and do the following:
• For each command type, find the last command of that type that was run for that ID. Save its index in the command list for later.
• For some command types, finding the last command is insufficient: for example move connector point needs the last start point move, and the last end point move. In these cases save both indexes.
• Move commands are a special case, since multiple types of commands perform absolute moves (move, connector move point, resize depending on args). Save the index of the last command that causes a move, and probably don't save the last move command earlier since this will cover that.
• Repeat for all element IDs. At the end, take the resulting set of indices, and construct a command sublist by taking only the commands at those indices in the command list.
• Save the command sublist.

Obviously this is a WIP solution and I don't know if it'll work. There are also parts of this to figure out: we'll likely need a way to figure out what elements an arbitrary command actually affects (casting might work, but look into interfaces for a cleaner solution?).