Improve LIFO and Vector Coordinator performance

[T.Doenz]

The LIFO and Vector coordinators are currently very slow (especially for take operations), compared to similar coordinators (FIFO, Key) and the old MozartSpaces version. The reason lies probably in the coordinator implementations, which should be tested, reviewed and improved. See the first benchmark results (100826_space_perftest.ppt) for detailed numbers.

[T.Doenz]

The LIFO performance has been improved with SVN rev. 10101. The Vector performance has been improved with SVN rev. 10104, but it is still slow. The main reason/line for its inferior performance needs further changes and is marked with a TODO comment

[T.Doenz]

Due to the refactoring in SVN rev. 10957 also the Vector coordinator should now be faster, but benchmarks need to be performed to confirm that.

[T.Doenz]

The changes in SVN rev. 10957 introduced a bug (the entry indices change when an entry at the beginning or in the middle is removed). The current version should be correct but has the same performance limitations as before.

[T.Doenz]

I close the ticket because the current coordinator performance is acceptable. The performance of LIFO is comparable to FIFO. The performance of Vector is lower but not easy to improve because of the coordinator requirements (direct access with index).