[Intel-gfx] forcewake patches (ignore previous, please)

[Ben Widawsky]

The request to test, and report the warnings is still desired.

These address all of Chris' comments from IRC (I think). I tried to put
the reasoning in the code comments, as well as commits; but I'll add it
here for completeness.

Simplify the acquisition of the forcewake lock by removing the forced
entry. The motivation behind it is that the kernel offers sufficient
tools to get relevant data before/during/after a hang, and adding a
complex mechanism for userspace is not such a great idea. If a need to
do this comes up, we can always add it back on a situational basis.

Similarly, on releasing of the debugfs file, no more forced entry mechanism.
This does leave a potential for release to hang, but AFAICS if release hangs,
the system is going to need a reboot before i915 becomes usable again because
module unload would also hang.

Ben