Metadata spec should be versioned

[TomAugspurger]

As mentioned in #89, we don't actually have "versions" anywhere in the spec. We should include something in the spec document (and any jsonschema if / when we provide that for the spec).

Additionally, should we consider adding a version field to the parquet metadata? That could have been helpful for the change proposed in #89 (it'll end up not being necessary if we just switch to stac:collections, but it's indicative of what we might want it for).

However, I recall that some systems have trouble using metadata so perhaps requiring that stac-geoparquet files have that set is too much.

[gadomski]

However, I recall that some systems have trouble using metadata so perhaps requiring that stac-geoparquet files have that set is too much.

We could put the requirement on writers, at least? Presumably stac-geoparquet writers should be sophisticated enough to write metadata, even if the consumers aren't? (FYI I preemptively added it to stac-rs here: stac-utils/rustac#677)

Suggested language:

stac-geoparquet files must contain a stac:geoparquet_version field that is a string of the file's stac-geoparquet version, e.g. "1.0.0".

👍🏼 to json-schema for the metadata expectations. Updated my stac-rs to conform to that idea: stac-utils/rustac#686

[TomAugspurger]

Presumably stac-geoparquet writers should be sophisticated enough to write metadata, even if the consumers aren't?

Yep, that's obvious in hindsight :)

[TomAugspurger]

Closed by #98