Handle VBT panel_type=0xff

Currently we fail to handle the VBT panel_type=0xff case. What that means is that we should read the EDID first, then match the PNPID from the EDID against the panel PNPIDs in the VBT LFP data block. The index of the matched entry should be our panel_type.

Roadblock 1: VBT parsing currently happens before output probing (and thus before the EDID is read). This needs to change to at least defer parsing of all panel_type dependent parts until output probing happens.

Roadblock 2: To read the EDID we need to turn on the panel VDD. This requires interacting with the panel power sequencer, and to do that we may need to initialize the panel power sequencer registers, which in turn depends on getting the panel power sequencing delays from VBT, which depends on panel_type. I do believe the GOP/VBIOS on most machines does leave the VDD override enabled after boot even if the panel is otherwise not enabled, so it should probably be possible to skip the VBT power sequencing delay stuff initially and just trust that whatever is in the registers at the time is sufficient. So we need to adjust the power sequencer init to skip the VBT stuff (at least if the panel power/VDD is already enabled), and make sure we don't turn off the VDD until after we've done the output probing (that may already be the case, not sure).

Roadblock 3: We actually need to do the PNPID match correctly. This one I've mostly typed up already: https://github.com/vsyrjala/linux.git vbt_panel_type_pnp_id

Several examples of EDID+VBT pairs that need this are attached to issue #125 (closed)