Add memory statistics for the global ResourceCache

[ptrgags]

Part of #9886, this PR tracks memory usage in the global ResourceCache used to load resources for ModelExperimental. It is the global counterpart to #10397 which tracks memory on a per-ModelExperimental basis.

Some important caveats:

• This only measures memory loaded through the cache, not all global memory usage in general.
• There is overlap between the memory counted here and the memory counted in Add basic memory statistics in ModelExperimental #10397. Memory can be shared between multiple tilesets due to the global cache. Meanwhile, a ModelExperimental may use a mix of resources loaded from the cache, and additional buffers generated at runtime as needed. Determining the size of the overlap is non-trivial.

Sandcastle for testing -- In the Logging section of the inspector, toggle on the new ResourceCacheStatistics like so:

@j9liu could you review?

CC @lilleyse

[cesium-concierge]

Thanks for the pull request @ptrgags!

• ✔️ Signed CLA found.
• ❔ CHANGES.md was not updated.
  • If this change updates the public API in any way, please add a bullet point to CHANGES.md.

Reviewers, don't forget to make sure that:

• Cesium Viewer works.
• Works in 2D/CV.

[j9liu]

@ptrgags did a first pass and left some comments!

[ptrgags]

Just leaving another Sandcastle link for testing property tables in the inspector, still have to fix the missing unit tests

[ptrgags]

@j9liu updated!

[j9liu]

@ptrgags -- did another pass. The biggest glaring issue is that the dropdown in the inspector isn't formatted well... maybe something additional needs to be modified in the HTML?

Also this PR should get logged in CHANGES.md.

[ptrgags]

@j9liu updated!

[j9liu]

@ptrgags you missed one of my comments; there's a unit test in ResourceCacheSpec, "loads index buffer from accessor as buffer", where the statistics aren't counted at the end of the unit test.

EDIT: also CHANGES.md

[ptrgags]

@j9liu while trying to fix the broken test, I learned that the metadata buffer views are copied and unloaded from the cache. The overall StructuralMetadataLoader is not cached, so really it shouldn't have been counted after all. I'll work on removing it from the count

[ptrgags]

@j9liu updated!

[j9liu]

@ptrgags - I just fixed a typo in the changelog, otherwise this looks good. I'll wait on Travis to pass before I merge.