[pull/stream] add a set of primitives used for moving data around

[tmpvar]

If the act of programming can be boiled down to moving data from one place to another, then it would make sense to have an abstraction that lets people reuse common data operations. This is a fundamental Lego brick in the rawkit toolset.

This is based on the Node.js pull-stream system, started by Dominic Tarr, but implemented in a C99 (C++11 compatible) way.

Note: this is a first round implementation and the contained streams will maybe become a bit more complex as monitoring/logging/similar bits are added, but the interfaces should be fairly solid.

At the core of this implementation is a ps_t (aka a "pull stream") to which a single function is bound on the fn property. For source streams, that is the only requirement. For sink and through streams another property must be populated before calling fn, the source property. source is a pointer to another ps_t and when hooked up it acts like two links in a chain or perhaps more c-esque, a linked list.

ps_t.fn has the signature ps_val_t * fn(ps_t *s, ps_stream_status status); which effectively breaks down to:

pull a value from a stream

ps_t *val = stream->fn(stream, PS_OK)
// if stream->status is not PS_OK then we're done - we can pass PS_ERR or PS_DONE to _signal_ the stream
// val can be NULL if the stream did not have any data for us
// if val is not NULL, the caller now owns `val` and `val->data` which are
// both required to be `ps_destroy`able.

TODO

• add readme
• link exposed tests in rawkit-test (will require a rebase of this branch)