consider dropping unnecessary request information when storing in Cache API

Currently the Cache API spec is written such that the entire Request is stored as the key in a table. All information about the Request is preserved. In theory this allows callers to use cache.keys() to get requests back out and initiate a fetch with them.

In practice implementations do not preserve this information in a compatible way. Gecko has tended to try to preserve all values as indicated by the spec. Chromium currently only preserves the information necessary for matching; i.e. the url and headers.

Since getting requests back out of cache.keys() is a bit niche, do we still want to require storing all information on the request? For example, things like mode, credentials, destination, internal origin value, etc.

Start with the Cache API specification's Request storage and cache.keys() behavior; compare the Gecko and Chromium preservation described here. Done means deciding whether the specification should continue requiring full Request information and documenting the resulting interoperability requirement.