Made answer sound more human like by using the propositions, adding a 'the' and using the name of the location not the object belonging to the location

- used general home directory as default ltm_base_path
- catched expception when certain needed memories where not exported
- reduced verbosity further
- removed some unnecessary comments

Merge branch 'feature/episodicVerbalizationSkill' of git.h2t.iar.kit.edu:sw/armarx/skills/speech into feature/episodicVerbalizationSkill

- added pythonproxy to the skill provider for episodic verbalization
- added parameters for LTM to object last seen at skill
- changed ice interface for querying symbolic scene memory
- calling python script for querying symbolic scene memory now
- changed timeout for object_last_seen_at skill to 1 minute

Merge branch 'feature/episodicVerbalizationSkill' of git.h2t.iar.kit.edu:sw/armarx/skills/speech into feature/episodicVerbalizationSkill

Since this is not done in advance but when constructing the API object, it only slows down all QA requests, including the ones that wouldn't use search at all (which applies to the most requests).

added ice interface to trigger memory search for symbolic scene memory and python script to search the symbolic scene memory

changed sleep duration
removed calls for tree creation and update from skill itself