Fix inverted layers in seeding

While porting the HLT1 seeding logic to HLT2 I also copied over the HLT1 layer order which is slightly different from HLT2. Here is the difference between the 2 orders on MC:

Blue is last MR, red is this one, green is before last MR. The order in this MR has a minimal impact on efficiency but do further reduce ghosts.

Also, here is a comparison on PbPb data (thanks @leuecker):

In these plots, blue is before last MR, yellow is last MR and green this MR.