Multithreaded hostcall support

Currently, a single thread serves hostcalls, but it should be possible to spawn multiple threads to service the same hostcall (especially once #50 is merged) concurrently. This would be useful for processing pipelines running on the GPU, where we want to minimize the amount of time spent waiting on the CPU to service the request. We would need to add parallel signals and buffers, and then have some scheme by which the GPU chooses which CPU thread to communicate with.

Start by locating the existing single-threaded hostcall implementation and review issue #50 for related changes. Define how parallel signals and buffers should support multiple CPU service threads and how the GPU selects among them. Done means the same hostcall can be serviced concurrently by multiple threads, with tests or benchmarks demonstrating concurrent processing.