Fix a bug where isActive could be seen to be different between engine and user access

[tiennou]

This is about #140, where the engine does the isActive check one way and later, while in tick does it a different way, leading to a bunch of bugs where structures that should be active actually aren't and vice-versa.

This is a slightly modified version of #107, updated to work on the latest version. The gist of it is that instead of doing checkStructureAgainstController checks, those are all made as part of the room processing logic, all at once, and set an off property on the relevant objects (anything that's in STRUCTURE_CONTROLLER, as part of a new _check_active_structures intent. Then all uses of the old function are changed to look at that off property to decide whether they're actually active or not.

This has been tested on a private server, console-boosting a room up to RCL8 then checking progressive lowering of the level, and appears to work properly.

[tiennou]

As an extra data point, CPU usage of .isActive() is now around 0.08-0.13, so a change to the CPU cost in docs might be warranted.