degradation in throughput of 10 to 20% (on windows)

Instead of the doAppend method being synchronized, WriterAppender now uses
UnsynchronizedAppenderBase as its super class. Consequently, various converters
can no longer assume that they are invoked from within a synchronized block.
 

- minor refactoring of the code (mostly renaming and moving to different package)

- added remark about log4j no longer being actively developed

had trouble with the old names.

perform threaded computations.

This commit is related to LBCLASSIC-36 as well as LBCORE-97

LBCLASSIC-36 Test performance of Yoda time compared to
SimpleDateFormat.

Here are sample results on my machine:

Raw Joda:               average tick 2714.052123 nanoseconds
Sync SimpleDateFormat:  average tick 2129.857639 nanoseconds
Raw SimpleDateFormat:   average tick 2138.306659 nanoseconds

Conclusion: Yoda time and SimpleDataFormat have roughly equivalent
performace, with the latter having the edge.

- We don't need to interrupt in LockingInJava

LBCLASSIC-133  We don't need to unit test test. A standalone app, is just as good if not simpler to run.

  appropriate

runnable[0]: counter=1002
runnable[1]: counter=0
runnable[2]: counter=0
runnable[3]: counter=0
runnable[4]: counter=0 

on the context name. This fits nicely the requirements expressed in LBCLASSIC-117

data in the form of StackTraceElement[] instead of CallerData[].
CallerData class now only extracts caller data. It has no instance
variables.