drm/pancsf: Allow for all GPU interrupts to be shared via a common framework

Not all platforms have a dedicated interrupt line going into the GIC for each block of the GPU. On Juno, FPGAs only have one interrupt line, so the handlers have to share that.

While updating the code I've realised that there is a lot of commonality between the interrupt handlers used by various files, and where they differ they also contain omissions so I have decided to extract the generic parts into a common hard interrupt handler and a common threaded interrupt handler that will allow for behaviour customisation via a hook function. The overall change does end up with a net code increase, but I think we also gain consistency in the handling of interrupts as well as allowing for future cleanups.

Signed-off-by: Liviu Dudau liviu.dudau@arm.com