Bug fixes including choosers that don't appear and livewindow widgets that don't appear.

[chauser]

Several significant changes and two rather cosmetic changes.
Significant changes:

• Arranged that connection to the robot always occurs after processing the save file. Fixes the problem that Choosers don't show up if connection to the robot happens before a chooser in the save file is processed. Also probably related to issue Command buttons don't work when loaded from save file #81 which mentions the same problem with buttons.
• Add LiveWindow widgets to containing subsystem widget when creating them from the save file. Fixes issue Smart Dashboard loses data in Test Mode #73.
• Registering the SchedulerType properly allows the Scheduler widget to be instantiated when the robot code places the Scheduler on SmartDashboard with SmartDashboard.putData.
• NetworkTable.getTable returns a new table with its own set of entries so it should only be called once per table. Changed Robot.java to implement this behavior.
• Fix issue Cancel buttons in Scheduler widget cause exceptions #187. Unfortunately I can't see any way to fix Cancel command button in the Scheduler Widget cancels the command forever #188, given the way the WPILib Sendable for the Scheduler works on the robot and the limitations of NetworkTables version 3.

Cosmetic changes:

1. When SmartDashboard is run with its stdout/stderr visible it produces a lot of uncaught java error index errors and tracebacks for choosers before the set of choices has been loaded.
2. Similarly, when SmartDashboard does not have a LiveWindow widget for a device type (e.g. CANCoder and many others) it produces an uncaught null pointer error message and traceback.
   Although these errors are uncaught, they occur on single-use callback threads so they don't actually keep SmartDashboard from operating, but they are a major distraction when trying to work on actual problems. Resolves SmartDashboard does not support several LiveWindow types #184; SmartDashboard can't display DifferentialDrive or MecanumDrive in LiveWindow mode. #97 is related in that it mentions the stacktraces for unimplemented device types but it asks to have them implemented.

[sciencewhiz]

It feels like the disabling DS connection should be a separate PR, since that is a workaround that could hopefully be reenabled in the future.

[sciencewhiz]

It feels like disabling DS connection may also workaround #81

[chauser]

Yes. I've separated the DSClient change into a new branch and will make a new PR for it.

[chauser]

Yes. I've separated the DSClient change into a new branch and will make a new PR for it.

Changes to DSClient are back in this PR but as a delay in starting the DSClient until after the save file is processed, rather than completely disabling it. I think this will be satisfactory in providing both automatic connection and resolving the issue of things not appearing.