Generalise regex files handling
Caselists, skips and flakes are all represented by a single file whose lines are regexs. For Mesa CI, it is convenient to factor these lists into separate files, e.g. one for all drivers, one per manufacturer, and finally ones specifically for a certain model of GPU. Currently the skips generally applicable to a driver family are being duplicated in each GPU specific skip list, and some global skips are not being merged into specific lists. By allowing separate files to be specified for each of these cases, which are then combined, it reduces the effort needed to maintain these files.
See also (for motivation)
Note that baselines are not covered here, since there is a semantic question of duplicates containing differing expectations, and in which order they should be resolved. It might be useful to define that, but I have no immediate need myself.
Piglit's runner has some duplication of read_lines logic, I left it alone, since I am not confident to properly test Piglit's interface.
I tested this by doing local runs and checking --caselist --skips and --flakes were working as intended.