EXT_feature_metadata JSON schema cleanup

[kring]

Fixes some small things I noticed while adding this extension to cesium-native:

• Specify that when the min/max/etc properties are arrays they must be arrays of numbers, not arrays of just anything.
• Remove possibly invalid (and redundant anyway) not.required property.

[kring]

By the way, are we sure we want specify min/max/etc to be "type": ["number", "array"]? glTF accessors have min/max properties that are very similar, and in that case they're always arrays. If the accessor is a scalar, it's simply an array of one element. This is much easier to deal with in statically typed languages.

[lilleyse]

This is to ensure that schema and schemaUri are mutually exclusive and was copied from camera.schema.json in the glTF spec:

https://github.com/KhronosGroup/glTF/blob/c6a5521ebd1469aba076949a97f150bf3591b63f/specification/2.0/schema/camera.schema.json#L36-L38

I wonder if there's a better way to do this.

[lilleyse]

The type should "integer" in both places here. I just pushed a fix related to that b3f48c2

[lilleyse]

By the way, are we sure we want specify min/max/etc to be "type": ["number", "array"]? glTF accessors have min/max properties that are very similar, and in that case they're always arrays. If the accessor is a scalar, it's simply an array of one element. This is much easier to deal with in statically typed languages.

I've debated this too. I'm not a huge fan of single element arrays since scalar properties tend to be way more common in metadata and the array form adds clutter/confusion IMO. I'd rather see the min/max value match the type exactly. Also there is better symmetry with the "default" property which doesn't use array form for single elements either.

I think glTF gets away with it because scalar accessors are not all that common. I understand why it's done though.

[lilleyse]

@kring thanks for the fixes and sorry for such a late review on this, I missed have must it when you posted it. Looks good I just had a couple comments.

[lilleyse]

Merged these changes in #10