A significant portion of players who use ports are passing objects through them. Currently they are required to handle that themselves via JSON serialization. This PR adds better support for passing objects; which is more convenient, extensive, and optimized (probably, more on this one later).
This adds zero overhead to existing (or when passing any primitive types) port usage, and also isn't a breaking change. The questions to debate here are:
Should objects be supported in the first place?
If so, how exactly do we want to serialize objects?
Based on an extensive discussion in Discord, the overwhelming majority answered "yes" to question one. As for question two, that has been much more hotly contested.
Ultimately, `structuredClone` was used, despite less-than-stellar performance, because other options were worse either in safety, speed, error-handling, or multiple of the above.
* Types for InternalFunction and ExternalFunction have been modified to actually typecheck ns functions against docs.
* Internal functions are required to use unknown for any params on the inner function.
* Return types for internal function inner-function must match the respective external function.
* Added new typecheck assertion function for asserting dynamic object types, to allow unknownifying params that were previously hardcoded objec structures.
* Because type assertion for parameter types and return types is enforced by InternalAPI, removed all duplicate type declarations on NetscriptFunction params and returns.