Need to consider globals

[tidoust]

Reporting the gist of the discussion in w3c/webref#1142 (comment): to avoid redundancy, interfaces may not be flagged as [SecureContext] in specs when they are only exposed to globals that are themselves only available in secure contexts, such as the ServiceWorker global. As a result, the logic that determines whether an IDL construct is secure context only also needs to take globals into account.

Updating the logic may not be worth the hassle in practice. This would likely only add the ExtendableCookieChangeEvent to the list of interfaces that are only usable in secure contexts.

(For Service Workers, the underlying interface ServiceWorkerGlobalScope did not have a [SecureContext] flag. Now fixed in the spec!)

[wbamberg]

Thank you François! I agree we should add this check, and yay for having ServiceWorkerGlobalScope marked secure context!