longer latency

- Increase latency from 850ms to 950ms (this makes the tests pass
quite consistently on Linux)

instead of logback-examples

- fixed http://jira.qos.ch/browse/LBSITE-35

reduce the size of the distribution which as a consequence went from
20 MB to 3MB.

slf4j/integrator module

working with development done on Linux where the line endings is LF.
This is still somewhat experimental.

hence the reapperance of LBCORE-26. Some tests still fail. Ongoing
work.

thread at least once before the next rollover occurs

This is OK, since in production, the rollver period is much much
longer than the time it takes to do compression. In our tests however,
this reation does not hold.

The name of the log file nested within .zip archives was incorrectly
set to be that of the tmp file logback creates during
compression. Anders Wallgren reported and posted an appropriate patch
for this problem LBCORE-98.  The nested file is now named after the
zip archive file. Analysis showed that this was an appropriate
strategy in all the cases we considered.

suggested by DAvid Varnes in LBCORE-101

- Adapted TimeBasedRollingWithArchiveRemovalTest so that
dailySizeBasedRollover() no longer fails under certain circumstances,
e.g. when in the last period no size-based rollover occuers. Also
increased the number of ticks per period (more logging calls) and
increased the triggering size (less size-based rollovers for faster
test execution time)

reference/export slf4j by variable.