IF: Shutdown thread pool in controller destructor

[heifner]

After discussing #2185 with @greg7mdp decided to move the shutdown of the controller_impl thread pool to the controller destructor. The real issue here, I believe, is that we have a circular reference between controller <-> controller_impl. I think we should break that circular reference, but until we do, this seems like a much safer approach.

This PR takes a step in the direction of removing controller_impl::self. It moves the signals to controller_impl and provides accessors on controller. It also removes all internal controller uses of self except for passing it on to transaction_context. Changing transaction_context and apply_context to use controller_impl instead of controller is a bigger change left for later. This PR does make the breaking change of signals being accessors instead of direct members on controller. Since we have changed what many of the signals emit in 6.0, seems like a good time to make this interface change. The removal of usage of self in transaction_context and apply_context can happen later without impacting external users of controller.

Depends on: AntelopeIO/reference-contracts#50 for the libtester tests.

[ericpassmore]

Note:start
group: IF
category: INTERNALS
summary: Better management of controller_impl thread pool.
Note:end

[linh2931]

Nice clean up!

[greg7mdp]

can you remove the member controller& self; at line 793?

[heifner]

can you remove the member controller& self; at line 793?

Not without the refactor of transaction_context and apply_context as it is passed in on every transaction execution.