Don't ever auto-define NDEBUG.  Just decide KJ_DEBUG based on optimization mode if neither DEBUG nor NDEBUG is defined.

Extend totalSizeInWords() to also return a count of capabilities, which helps when a separate capability table needs to be allocated as well.  Use this in the RPC system.

Optimize promise tail calls by making ChainPromiseNode automatically detect and remove redundant nodes.

Change capability pointers to be indexes into a separate cap list so that CapDescriptors can be interpreted on receipt rather than delaying until the application actually traverses the message.  This massively simplifies a lot of things.

Get rid of 'Advanced' server methods, just have one method that takes a CallContext.  A typedef reduces boilerplate.

Make all Promise methods consistently consume the promise (returning a new promise when it makes sense), rename daemonize -> detach, and make eagerlyEvaluate() require an error handler (this caught several places where I forgot to use one).

Split AsyncIoProvider into high-level and low-level interfaces, so that alternate EventPort implementations can implement the low-level interface and reuse the higher-level stuff.

Overhaul the way EventLoop is specialized so that it's possible to hook up to an existing event loop infrastructure that is not KJ-aware.  This also makes the async IO API more dependency-injection-friendly.