aco: compact aco::span<T> to use uint16_t offset and size instead of pointer and size_t.

This reduces the size of the Instruction base class
from 40 bytes to 16 bytes. No pipelinedb changes.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3332>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3332>

src/amd/compiler/aco_instruction_selection.cpp

@@ -7964,7 +7964,7 @@ void split_arguments(isel_context *ctx, Pseudo_instruction *startpgm)
    /* Split all arguments except for the first (ring_offsets) and the last
     * (exec) so that the dead channels don't stay live throughout the program.
     */
-   for (unsigned i = 1; i < startpgm->definitions.size() - 1; i++) {
+   for (int i = 1; i < startpgm->definitions.size() - 1; i++) {
       if (startpgm->definitions[i].regClass().size() > 1) {
          emit_split_vector(ctx, startpgm->definitions[i].getTemp(),
                            startpgm->definitions[i].regClass().size());