- make closures know where they should be executed (eg, on a workqueue,
  or a combiner, or on an exec_ctx)
- this allows removal of a large number of trampoline functions that
  were appearing whenever we used combiners, and should allow for a much
  easier interface to combiner locks

IF:
- we schedule a write in chttp2 in response to some stream op
  (which will cause a write that's covered by a poller to be scheduled
  on the combiner lock)
- AND then, under that same combiner lock, we process a RST_STREAM
- then we'll remove the op that's being processed, consequently removing
  the polling coverage
- and then, IF that is the last poll on said transport, the transport
  will never write, which CAN cause servers to fail to shutdown

Allows exec_ctx callbacks to be called while a combiner is executing.
Also allows guaranteeing direct execution of callbacks from combiners,
which should allow reducing cpu burn for up/down stack interactions in
the future.

Afford other threads the opportunity to insert into a combiner prior to the final list being executed but after the async pause

Rename async_execution_lock --> combiner, make it use closures, unroll previous changes for idleness

This reverts commit c9d4b81d.