[Intel-gfx] Lock-correct batch-and-ring-buffer dumping

[Carl Worth]

Here's a refresh of code originally written by Ben Gamari for
debugging batch-buffer content.

My new contribution is the first commit which adds a spinlock
to protect the active list. Then Ben's code is changed to use
this spinlock rather than grabbing the struct_mutex. The idea
here is to allow the batch-buffer to be dumped even if things
are hung with the struct_mutex held.

Ben, can we get a Signed-off-by from you for these?

Thanks,

-Carl