Model Loading optimization possibility

[vade]

Today in Satin, model loading can make many nested transforms, mirroring USD graphs etc

We should optionally compute nested transforms and only add Meshes / Submeshes at the top most level, simplifying the recursion needed in methods like updateLists, which for large models, dominate the CPU usage and makes folks sad

of 1297 objects only 271 are actual meshes, which could have had their root transforms updated and avoided 81% of recursion / cost in updateLists

po object.children[1].getChildren(true)
▿ 1297 elements
  ▿ 0 : <Object: 0x1229f7a00>
  ▿ 1 : <Object: 0x122a3e000>
  ▿ 2 : <Object: 0x122a45a00>
  ▿ 3 : <Object: 0x122a46000>
  ▿ 4 : <Object: 0x122a92800>
  ▿ 5 : <Object: 0x122a92e00>
  ▿ 6 : <Object: 0x122a6d000>
  ▿ 7 : <Object: 0x122a6d600>
  ▿ 8 : <Object: 0x12298de00>
  ▿ 9 : <Object: 0x12298e400>
  ▿ 10 : <Object: 0x12298ea00>
  ▿ 11 : <Object: 0x122870a00>
  ▿ 12 : <Object: 0x122871000>
  ▿ 13 : <Object: 0x122871600>
  ▿ 14 : <Object: 0x12286ec00>
  ▿ 15 : <Object: 0x12286f200>
  ▿ 16 : <Object: 0x12286f800>
  ▿ 17 : <Object: 0x122a49000>
  ▿ 18 : <Mesh: 0x122a49600>
  ▿ 19 : <Object: 0x12384be00>
  ▿ 20 : <Object: 0x12384ce00>
  ▿ 21 : <Object: 0x123811200>
  ▿ 22 : <Mesh: 0x12387f600>
  ▿ 23 : <Object: 0x122072c00>

[vade]

This is mocked up in Fabric and seems to work, ill port it into Satin