MMC none blocking requests

perfor 2011-06-13: Postpones 3 WIs because the SDIO investigation seems to be enough to motivate new function in core.c for async request handling. There seems to be no major adjustments needed in order for an SDIO client to use the async request handling in core.c. Adding async support to the SDIO fwk will be covered by https://blueprints.edge.launchpad.net/linux-linaro/+spec/sdio-async-request.
perfor 2011-07-11: Postpones "start from interrupt context" because it's suboptimal. Refer to https://lkml.org/lkml/2011/6/30/230.

Work items:
[per-forlin] implement async request handling in core.c: DONE
[per-forlin] run performance test and liability test to verify implementation: DONE
[per-forlin] post patch async req patch on mmc-list: DONE
[per-forlin] Set up SDIO client for test purposes, wl12xx on pandaboard: DONE
[per-forlin] investigate if/how SDIO clients may utilize async request support in core.c: DONE
[per-forlin] rebase patches on top of linux 3.0, run all tests: DONE
[per-forlin] set up and run w12xx_sdio_test on pandaboard (or any other suitable board): POSTPONED
[per-forlin] implement test_case in w12xx_sdio works as a reference implementation for async req for SDIO: POSTPONED
[per-forlin] run performance test to measure SDIO bandwith gain with performance and ondemand governor: POSTPONED
[per-forlin] post linux 3.0 mmc async req support on mmc-list: DONE
[per-forlin] linux 3.0 mmc async req, update after review: DONE
[per-forlin] mmc-next linux 3.1 mmc async req, rebase and test: DONE
[per-forlin] mmc-next linux 3.1 mmc async req, submit and review: DONE
[per-forlin] Investigate if/how to split up large single request in order to minimize prepare latency: DONE
[per-forlin] Investigate if/how to start next async req from interrupt context instead of wait_for_completion: POSTPONED