mesa/dlist: reduce CPU overhead
- glthread merging of successive glCallList in a single glCallLists
- less locking
- cleanup / removal of unused code
- cache misses reduction
- less branching
(this MR includes the idalloc commits from !10860 (closed); the util/idalloc: add util_idalloc_alloc_range
commit has been modified to fix bugs in the original commit)
Improve SPECviewperf 13 snx test performance with RADEON_NOOP=1 by 10%-100% depending on the subtest (overall "composite score" increase: 35%).