Optimize name-id-lookup for mapblock serialization

[Desour]

• Goal: Improve performance of mapblock de-/serialization.
• Commits should be self-explanatory.
• Interesting side-node: It looks to me like loops over all nodes in a mapblock are completely rolled out. Idk if this is fine or not. Edit: Oops, found the jmp now.

To do

This PR is a Ready for Review.

How to test

• time luanti --server --recompress --worldname name_of_your_world
• (For reproducibility, first do it once to make sure the blocks are all in newest format, and avoid differences due to caching of the db file.)
• Tested with a ca. 100MB block data db:
  • master:
    • real 0m52,698s
      user 0m43,924s
      sys 0m2,004s
    • real 0m50,189s
      user 0m44,164s
      sys 0m1,665s
    • real 0m57,698s
      user 0m50,291s
      sys 0m1,936s
  • PR:
    • real 0m44,240s
      user 0m35,712s
      sys 0m1,894s
    • real 0m42,180s
      user 0m36,517s
      sys 0m1,686s
    • real 0m42,680s
      user 0m37,303s
      sys 0m1,793s

[sfan5]

LGTM