Increased timeout of skill RandomChaining to 2 minutes

This reverts commit cb37cc10.

Improved locking of subskillsMutex during callSubskillAsync to avoid possible deadlock during abortion of skills

Loading the last n Snapshots from LTM on Startup

See merge request !469

Diagnostics sub-unit with battery management

See merge request !471

- all Memories can now be used to reload recorded data on startup (tested with Vision, RobotState and ObjectMemory)
	- VisionMemory seems to override the loaded snapshots, this needs more investigation

- properties are part of LTM for now, which is ugly but the only easy way for now to do it

Implemented loading only the newest n snapshots from the given CoreSegments (String has to match core segment name from id)
- n and the names are currently hard-coded but need to be loaded from properties
-

- names are still hard coded at the moment

Currently all snapshots are loaded from all core segments and all days of the memory server.
The method gets called by the memory server component itself.

Revert "fix for recording with LTM when dealing with broken "sent"-timestamps, like in Proprioception of RobotState Memory."

This reverts commit c85eac75.

As the snapshots are the oldest though, they will be forgotten relatively fast if maxHistorySize for that CoreSegment is small.
Currently import path is assumed to be the same as export path or empty.
When empty, nothing is done, when non-empty export path and export name are used to import last snapshots for this memory.

fix for recording with LTM when dealing with broken "sent"-timestamps, like in Proprioception of RobotState Memory.
In this case the frequency filter takes the "arrived" timestamp, as this is normally not broken.

maxHistorySizeInfo reduced spamming

See merge request !467

tested in Simulation (also with multiple GUI instances)