Host Instance Reset Capability

[VistianOpenSource]

The current error policy is one whereby two differing methods are specified for the handling of errors. One of these returns an IPrimitive , the other is for a 'fatal' error during the commit phase.

What we don't current have however is the ability to 'reset' and start again.

So a 'reset' capability needs to be added to base implementation of host instances AND, the way in which the error policy is used needs to reflect this. It should be remembered that the current implementation of the layout uses Reactive Extensions and currently the way the error policy ia used is pretty prescriptive.

[VistianOpenSource]

Made changes to build host instance to now have a separation between the production of the layout and the actual build host instance. This allows for the swapping in of a different layout host for error conditions.

This meant loosing reactive extensions as being the basis for the host, handling errors and retries just became too difficult. Instead TPL dataflows were used to provide the basis for the layout host.