Cleanup and bugfix

• Catching write errors and emitting them
• Ensuring a stream is closed before finishing a join (internal testing method)
• Handling events in the underlying stream

Predictable rollover and non-raw stream support

• Support non-raw streams. For some reason, raw streams are MUCH faster in high load scenarios (at least when this is the only stream).
• Better guarantees over file rollover - we will write exactly one log record that goes over the size threshold before we rotate
  The previous performance release meant that we couldn't rotate until the write had completed to the disk - in the meantime several other
  logs could have been written. This made everything unpredictable.
• Making better use of the cargo datatype to write multiple log records in a single event loop tick.
• Using setImmediate rather than process.nextTick in the write loop to allow io and other operations time to happen rather than hog the event loop.

Performance

Checked performance of the log writing framework and tweaked a few things to improve this.

Bugfix - reenabling special values for periods

v1.0.6

Preparing v1.0.6

First release

Moved to a new repository that better matched the release.
Restructured the project and added some tests.